{ "abstract": "Introduction Therapeutic inertia (failure to initiate or intensify therapy when therapeutic goals are unmet) contributes to poor glycemic control and diabetes-related complications. We assessed the extent of therapeutic inertia, defined as a lack of new prescription orders for sodium-glucose cotransporter-2 inhibitors (SGLT2i) or glucagon-like peptide-1 receptor agonists (GLP-1RA), among patients with type 2 diabetes and above-target hemoglobin A1c who had clinical indications for use, were not currently using these medications, and had no contraindications. We also examined whether prescribing patterns differed by race and ethnicity.Research design and methods We conducted a retrospective cohort analysis of 2018–2022 electronic health record data. Log-link Poisson generalized estimating equation models estimated relative risks (RR) and assessed predictors of therapeutic inertia and trends over time.Results Among 10 345 eligible patients (30 740 encounters), 56% were White, 37% Black, 3% Hispanic, 2% Asian, and 2% other races and ethnicities. The rate of new prescribing increased over time but remained low (SGLT2i: 0.9%–4.3%; GLP-1RA: 1.2%–5.2%). After adjusting for sociodemographic, clinical, and service factors, Black patients were less likely than White patients to receive new SGLT2i (RR 0.59, 95% CI 0.42 to 0.82) and GLP-1RA (RR 0.62, 95% CI 0.49 to 0.79) prescriptions; Hispanic patients were less likely to receive new GLP-1RA prescriptions (RR 0.48, 95% CI 0.25 to 0.92).Conclusion Despite compelling indications, initiation of SGLT2i and GLP-1RA remained low overall and was significantly lower among Black and Hispanic patients, underscoring therapeutic inertia as a modifiable barrier to optimal diabetes care.", "authors": [ { "affiliations": [ "Department of Medicine, Duke University School of Medicine, Durham, North Carolina, USA" ], "name": "Jashalynn German" }, { "affiliations": [ "Department of Biostatistics and Bioinformatics, Duke University School of Medicine, Durham, North Carolina, USA" ], "name": "Wei Angel Huang" }, { "affiliations": [ "Department of Biostatistics and Bioinformatics, Duke University School of Medicine, Durham, North Carolina, USA" ], "name": "Amanda Brucker" }, { "affiliations": [ "Duke Health Integrated Practice, Duke Health, Durham, North Carolina, USA" ], "name": "Khayla Daniel" }, { "affiliations": [ "Department of Medicine, Duke University School of Medicine, Durham, North Carolina, USA", "Duke Primary Care, Duke University Medical Center, Durham, North Carolina, USA" ], "name": "David Halpern" }, { "affiliations": [ "Department of Surgery, Duke University School of Medicine, Durham, North Carolina, USA" ], "name": "Nrupen Bhavsar" }, { "affiliations": [ "Duke Health Technology Solutions, Duke University Health System Inc, Durham, North Carolina, USA" ], "name": "Eugenia McPeek Hinz" }, { "affiliations": [ "Duke University Health System, Durham, North Carolina, USA" ], "name": "Richard Shannon" }, { "affiliations": [ "Department of Medicine, Duke University School of Medicine, Durham, North Carolina, USA" ], "name": "Michael Pignone" }, { "affiliations": [ "Department of Medicine, Duke University School of Medicine, Durham, North Carolina, USA" ], "name": "Matthew J Crowley" }, { "affiliations": [ "Medicine, Division of Endocrinology, Duke University, Durham, North Carolina, USA" ], "name": "Bryan C Batch" }, { "affiliations": [ "Duke University School of Medicine, Durham, North Carolina, USA" ], "name": "Benjamin A Goldstein" }, { "affiliations": [ "Department of Medicine, Duke University School of Medicine, Durham, North Carolina, USA" ], "name": "Susan Elizabeth Spratt" } ], "full_text": "WHAT IS ALREADY KNOWN ON THIS TOPIC Sodium-glucose cotransporter-2 inhibitors (SGLT2i) and glucagon-like peptide-1 receptor agonists (GLP-1RA) reduce cardiovascular and renal risk in type 2 diabetes but remain underprescribed, partly due to therapeutic inertia. The degree to which this inertia varies by race and ethnicity is unclear.WHAT THIS STUDY ADDS In a large health system cohort with above-target A1c and clear indications, initiation of SGLT2i and GLP-1RA increased from 2018 to 2022 but remained low. Black patients were less likely to receive either drug, and Hispanic patients were less likely to receive GLP-1RA. Disparities persisted for GLP-1RA but attenuated for SGLT2i.HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY By quantifying persistent therapeutic inertia in prescribing, this study identifies measurable targets for intervention and establishes a baseline to evaluate efforts to increase adoption of SGLT2i and GLP-1RA. The findings can inform strategies to improve prescribing practices, expand access, and advocate for affordability to ensure equitable use of these medications.Introduction Diabetes mellitus affects approximately 37 million individuals in the USA (11% of the US population), and prevalence is expected to surpass 50 million by 2030. 1 2 Diabetes increases the risk of cardiovascular disease, kidney disease, and early mortality, with a disproportionate impact on socioeconomically disadvantaged and racial and ethnic minority groups.3–5 Over the past decade, the Food and Drug Administration has approved multiple agents from the sodium-glucose cotransporter-2 inhibitor (SGLT2i) and glucagon-like peptide-1 receptor agonist (GLP-1RA) medication classes. SGLT2i and GLP-1RA medication classes have collectively been shown to improve cardiovascular outcomes in type 2 diabetes, while SGLT2i have additionally been proven to improve renal outcomes.6–12 Since 2018, the American Diabetes Association (ADA) has recommended the consideration of these agents for patients with type 2 diabetes and known cardiorenal disease, as well as those at high risk for atherosclerotic cardiovascular disease or chronic kidney disease (CKD).13 In 2022, the ADA began recommending the consideration of prescribing SGLT2i and/or GLP-1RA as front-line agents for patients with type 2 diabetes and heart failure, established cardiovascular disease, or multiple risk factors for cardiovascular disease, irrespective of metformin use.14–19Despite the demonstrated benefits of SGLT2i and GLP-1RA, studies have reported consistent underutilization.15–21 Therapeutic inertia, defined as failure to initiate or intensify therapy when therapeutic goals are not met,22 is associated with prolonged hyperglycemia, higher risk of complications, and reduced life expectancy.23 24 Therapeutic inertia is influenced by complex interactions between factors at the patient (eg, fear of side effects of medication changes, cost, and pill burden), clinician (eg, concerns about medication cost and difficulty navigating guidelines), and healthcare system (eg, lack of decision support and limited team approach to care) levels.23 Racial and ethnic disparities in uptake of these agents have been documented,25–29 yet less is known about encounter-level therapeutic inertia at the point of care among eligible patients after accounting for access-to-care and utilization factors. This study addresses this gap by examining prescribing trends of SGLT2i and GLP-1RA among patients with type 2 diabetes, hemoglobin A1c (HbA1c) ≥8%, who met the clinical indications for these medications, focusing on both overall use and racial and ethnic differences in therapeutic inertia.Research design and methods Environment For this single-center retrospective cohort study, we used electronic health records (EHR) data from the Duke University Health System (DUHS) to construct the cohort, derive patient characteristics, and identify outcomes. DUHS consists of three hospitals and over 400 outpatient clinics that have used a single EHR system since 2014. As one of the main health systems in Durham, North Carolina, 30 a racially and ethnically diverse city and county, it provides insight into healthcare delivery across a broad patient population.Cohort Through our epic-based EHR system, we constructed a comprehensive diabetes registry that includes individuals with either an active problem list diagnosis of all diabetes types (International Classification of Diseases, 10th revision (ICD-10) codes), two health system encounters in the past 2 years (730 days) associated with a billing diagnosis of diabetes, or the presence of an antihyperglycemic agent on the medication list (with the exception of metformin or GLP-1 monotherapy regimens). Additionally, to be labeled as ‘active’ in the registry, an individual must be alive and have had an encounter in the past 3 years or be in the accountable care organization registry or have a scheduled appointment within the upcoming 6 months. Notably, patients cannot concurrently have a diagnosis of prediabetes on the problem list to be considered active in the diabetes registry. Our analytic cohort was restricted to adult patients (aged 18 years or older as of January 1, 2018) with type 2 diabetes who were active in the diabetes registry between January 1, 2018 and December 31, 2022. Encounters occurring in pediatric clinic settings represented a small proportion of the analytic cohort (440 of 30 740 eligible encounters, <2%) and occurred exclusively in general pediatrics; no encounters occurred in pediatric endocrinology clinics.Eligible encounters We identified patient encounters where diabetes treatment escalation was indicated. The analysis cohort consisted of encounters with patients who had compelling indications for SGLT2i or GLP-1RA therapy (based on ICD-10 diagnosis codes), were not currently prescribed either medication at the start of the encounter, and had no documented contraindications. Compelling clinical indications for prescribing medications of interest were defined as follows: history of ischemic heart disease, stroke, or diabetic renal disease (for both medications, informed by ADA guideline recommendations and secondary renal outcomes data available during the study period), congestive heart failure (for SGLT2i), or peripheral vascular disease (for GLP-1RA). Contraindications were specified as follows: a diagnosis of alcohol use disorder (for both medications); end-stage renal disease, diabetic ketoacidosis, osteoporosis, recurrent urinary tract infection, foot ulcers or lower extremity amputation (for SGLT2i); gastroparesis or pancreatitis (for GLP-1RA). Comorbid conditions were identified using ICD-10 diagnosis codes and grouped using Elixhauser comorbidity definitions ( online supplemental Appendix S1). Encounters were excluded if patients had an ICD-10 diagnosis code for type 1 diabetes, were pregnant during the study period, or had a recent history of solid organ transplant. Eligible encounters were defined as outpatient encounters with departments that routinely provide diabetes care (ie, Endocrinology, Internal Medicine, Family Medicine, Pediatrics, General Internal Medicine, or Pediatric Endocrinology) that were associated with a HbA1c ≥8% recorded during or in close proximity (±10 days) to this outpatient encounter. In sensitivity analyses, we restricted HbA1c values to those obtained prior to the qualifying encounter to ensure temporal alignment with prescribing decisions. Encounters in pediatric or pediatric endocrinology clinics were included only if they involved adult patients (eg, young adults receiving transitional care). No encounters involving patients younger than 18 years were included in the analysis. The analysis cohort included eligible patients who had at least one qualifying encounter in the study window (figure 1). To assess outpatient engagement, we examined visit history prior to the qualifying encounter and found that 88.7% of encounters were preceded by at least one outpatient visit within the prior 6 months, and 75.5% were preceded by more than one visit, indicating that most patients in the analytic cohort were actively engaged in outpatient care.SP110.1136/bmjdrc-2025-005546.supp1Supplementary dataFigure 1The cohort selection process from the Duke University Health System diabetes registry between January 1, 2018 and December 31, 2022. Of 200 720 adult patients with diabetes, patients were excluded based on clinical or data-related criteria. The final analytic sample included 26 835 patients with measured HbA1c values and at least one qualifying outpatient diabetes-related encounter. Two analytic subgroups were derived: one for GLP-1RA therapeutic inertia and one for SGLT2i therapeutic inertia. Each subgroup included encounters where the patient had a compelling clinical indication and was not already on the respective medication class. *An HbA1c ≥8% needs to be taken during or proximal to (±10 days) a qualifying outpatient encounter in Endocrinology, Internal Medicine, Family Medicine, Pediatrics, General Internal Medicine, or Pediatric Endocrinology. GLP-1RA, glucagon-like peptide-1 receptor agonists; HbA1c, hemoglobin A1c; SGLT2i, sodium-glucose cotransporter-2 inhibitors.Outcomes For each eligible encounter, we assessed whether there was a new SGLT2i or GLP-1RA prescription order within 14 days, analyzed separately for SGLT2i and GLP-1RA prescriptions. A medication prescription was considered active for 1 year after the order date if no end date was indicated in the EHR. Therapeutic inertia in the new SGLT2i and GLP-1RA prescriptions was defined as the absence of a prescription order following an eligible encounter.Comparators of interest Our primary independent variable was patient race and ethnicity, as recorded in the EHR. Categories included non-Hispanic Black, non-Hispanic White, non-Hispanic Asian, Hispanic, and other. Race and ethnicity were analyzed due to their well-documented association with inequities in diabetes prevalence and outcomes. 4 31 Our secondary comparator of interest was year to capture changes in prescribing over time.Additional patient and clinical characteristics Patient characteristics were defined at the time of each eligible encounter ( online supplemental table 1). Characteristics included individual and neighborhood-level sociodemographic variables, including age, sex, insurance status, and area deprivation index (ADI). ADI is a widely used, publicly available, composite measure incorporating income, employment, education, and poverty levels to establish state-level ranks of census block groups, assigning values from 1 to 10 for each area (with 10 being the highest level of disadvantage).32 Additional patient characteristics included a custom list of comorbidities of interest, vital signs, laboratory measures, and indicators of access to care (eg, number of eligible encounters, number of emergency room visits, hospitalizations, and outpatient visits with primary care physicians and endocrinologists, respectively, within the past 6 months before the eligible encounter). These variables were selected to allow for a more comprehensive assessment of potential mediators and confounders in the association between race and ethnicity, year, and therapeutic inertia.Statistical analysis We summarized patient-level baseline sociodemographic and clinical characteristics overall and stratified by race and ethnicity. For time-varying characteristics (age, lab values, and vitals), we used the patient’s status at their earliest eligible encounter. During the study period, estimated glomerular filtration rate (eGFR) reporting reflected institutional clinical practice, with implementation of the 2021 race-free Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation in March 2022. Differences across race and ethnicity subgroups were quantified using standardized mean differences (SMD), with SMD values of 0.2, 0.5, and 0.8 considered to represent small, medium, and large effect sizes, respectively. 33We assessed encounter-level therapeutic inertia separately for SGLT2i and GLP-1RA prescriptions. We calculated the absolute rate of medication prescriptions annually and stratified it by race and ethnicity. The prescription rate was calculated as the ratio of encounters with a new prescription over the number of eligible encounters (as defined above). We modeled the relative risks (RR) in a log-link Poisson regression framework using generalized estimating equations (GEE) with robust SE and first-order autoregressive correlation structure to account for repeated encounters within patients. Inference from each model focused on the primary comparisons of interest across race and ethnicity and year. We fit four nested models for each outcome and comparator variable: first unadjusted, then adjusted for sociodemographic, clinical, and service utilization factors cumulatively. This approach was employed to examine the potential mediating effects of these factors in the association of race and ethnicity and year with therapeutic inertia. We accounted for non-linearity in continuous variables using natural cubic splines. The interaction between race and ethnicity and year was additionally examined in an exploratory analysis. We treated missing values as a separate category when feasible (online supplemental table 2). RR and 95% CIs were reported from all model fits.We further examined time trends in prescription rates by race and ethnicity using the same GEE model framework. The encounter date was modeled flexibly as a natural cubic spline. We fit unadjusted models and fully adjusted models with sociodemographic, clinical, and service factors. We obtained predicted prescription rates at the means of continuous covariates and modes of categorical covariates and plotted 95% prediction intervals as shades.Sensitivity analysis We repeated the primary models, restricting HbA1c measurements to those obtained prior to the qualifying encounter. Because eGFR thresholds for SGLT2i initiation evolved during the study period, we conducted a sensitivity analysis applying time-specific eGFR exclusion criteria. Encounters prior to 2021 were excluded if eGFR was <45 mL/min/1.73 m², and encounters from 2021 onward were excluded if eGFR was <30 mL/min/1.73 m². Results from this analysis were compared with the primary analysis to assess the robustness of associations between race and ethnicity and therapeutic inertia.Statistical significance was defined as p-values <0.05 or 95% CI not crossing 1. No adjustments for multiple testing were performed. All analyses were performed in R V.4.1.3.This work was approved as exempt by the DUHS Institutional Review Board (Pro00111586).Data and resource availability The datasets generated and analyzed during the current study are not publicly available due to the inclusion of protected health information. Analytic code is available from the corresponding author upon request.Results Table 1 and online supplemental table 2 show baseline characteristics of the study population stratified by race and ethnicity.Table 1Patient characteristics at first eligible encounter: overall and by race and ethnicityAll patientsN=10 345NH White N=5832NH Black N=3820NH Asian N=228Hispanic N=274Other N=191Standardized mean differenceSociodemographic Age, median (Q1, Q3)67 (58, 74)68 (60, 75)64 (55, 72)68 (58, 74)62 (53, 70)64 (56, 72)0.236 Male sex, n (%)5464 (52.8)3461 (59.3)1609 (42.1)139 (61.0)152 (55.5)103 (53.9)0.175 Primary payer, n (%)0.377  Commercial3956 (38.2)2360 (40.5)1298 (34.0)112 (49.1)113 (41.2)73 (38.2)  Medicaid867 (8.4)254 (4.4)522 (13.7)27 (11.8)41 (15.0)23 (12.0)  Medicare2077 (20.1)1259 (21.6)705 (18.5)36 (15.8)39 (14.2)38 (19.9)  Medicare advantage2966 (28.7)1724 (29.6)1105 (28.9)42 (18.4)45 (16.4)50 (26.2)  Other182 (1.8)105 (1.8)62 (1.6)3 (1.3)7 (2.6)5 (2.6)  Uninsured297 (2.9)130 (2.2)128 (3.4)8 (3.5)29 (10.6)2 (1.0) Neighborhood area deprivation index, n (%)0.629  1–2 (least deprived)2757 (26.7)1718 (29.5)703 (18.4)172 (75.4)85 (31.0)79 (41.4)  3–42745 (26.5)1681 (28.8)915 (24.0)29 (12.7)73 (26.6)47 (24.6)  5–61878 (18.2)1110 (19.0)689 (18.0)11 (4.8)46 (16.8)22 (11.5)  7–81524 (14.7)752 (12.9)707 (18.5)6 (2.6)39 (14.2)20 (10.5)  9–10 (most deprived)1272 (12.3)502 (8.6)725 (19.0)2 (0.9)24 (8.8)19 (9.9)  Missing169 (1.6)69 (1.2)81 (2.1)8 (3.5)7 (2.6)4 (2.1)Laboratory and clinical measurements Hemoglobin A1c (%), median (Q1, Q3)8.7 (8.3, 9.8)8.6 (8.2, 9.5)8.9 (8.3, 10.3)8.6 (8.2, 9.4)9.1 (8.4, 10.7)8.7 (8.3, 10.0)0.252 BMI (kg/m2), median (Q1, Q3)31.5 (27.4, 36.6)31.5 (27.5, 36.2)32.0 (27.7, 37.5)26.1 (22.8, 29.0)30.0 (27.0, 35.3)31.3 (26.7, 37.1)0.435 BMI missing91 (0.9)54 (0.9)34 (0.9)1 (0.4)2 (0.7)0 (0.0) Estimated glomerular filtration rate (mL/min), median (Q1, Q3)60 (48, 71)60 (48, 72)60 (48, 67)60 (53, 74)60 (52, 81)60 (48, 72) Estimated glomerular filtration rate (mL/min), n (%)0.192  Normal (≥60)5907 (57.1)3274 (56.1)2193 (57.4)146 (64.0)184 (67.2)110 (57.6)  Moderate CKD (30–59)3411 (33.0)2045 (35.1)1190 (31.2)63 (27.6)52 (19.0)61 (31.9)  Severe and end-stage CKD (≤29)660 (6.4)314 (5.4)293 (7.7)13 (5.7)27 (9.9)13 (6.8)  Not measured367 (3.5)199 (3.4)144 (3.8)6 (2.6)11 (4.0)7 (3.7) Microalbumin/creatinine ratio (mg/g), median (Q1, Q3)30.00 (16.6, 148.1)30.00 (16.10, 84.12)30.00 (18.5, 300.0)30.90 (15.9, 269.1)30.00 (13.0, 178.2)30.00 (18.2, 207.2) Microalbumin/creatinine ratio (mg/g), n (%)0.17  Normal <302562 (24.8)1480 (25.4)909 (23.8)55 (24.1)72 (26.3)46 (24.1)  High 31–200764 (7.4)416 (7.1)291 (7.6)24 (10.5)19 (6.9)14 (7.3)  Very high 201–300591 (5.7)281 (4.8)266 (7.0)20 (8.8)11 (4.0)13 (6.8)  Extremely high >300388 (3.8)145 (2.5)205 (5.4)12 (5.3)18 (6.6)8 (4.2)  Not measured6040 (58.4)3510 (60.2)2149 (56.3)117 (51.3)154 (56.2)110 (57.6)Comorbidities Hypertension9467 (91.5)5272 (90.4)3612 (94.6)195 (85.5)223 (81.4)165 (86.4)0.197 Ischemic heart disease2943 (28.4)1905 (32.7)871 (22.8)61 (26.8)55 (20.1)51 (26.7)0.134 Stroke1335 (12.9)698 (12.0)564 (14.8)23 (10.1)31 (11.3)19 (9.9)0.071 Diabetic renal disease5318 (51.4)2796 (47.9)2168 (56.8)121 (53.1)144 (52.6)89 (46.6)0.102 End-stage renal disease240 (2.3)73 (1.3)147 (3.8)3 (1.3)14 (5.1)3 (1.6)0.122 Diabetic retinopathy1276 (12.3)517 (8.9)641 (16.8)42 (18.4)48 (17.5)28 (14.7)0.128 Gastroparesis158 (1.5)72 (1.2)75 (2.0)0 (0.0)7 (2.6)4 (2.1)0.11 Peripheral vascular disease2785 (26.9)1694 (29.0)935 (24.5)43 (18.9)66 (24.1)47 (24.6)0.098 Neuropathy4920 (47.6)2802 (48.0)1822 (47.7)86 (37.7)125 (45.6)85 (44.5)0.097 Diabetic ketoacidosis77 (0.7)26 (0.4)49 (1.3)0 (0.0)2 (0.7)0 (0.0)0.094 Alcohol use disorder0 (0)0 (0)0 (0)0 (0)0 (0)0 (0)– Pancreatitis213 (2.1)111 (1.9)91 (2.4)2 (0.9)3 (1.1)6 (3.1)0.086 Congestive heart failure2045 (19.8)1072 (18.4)868 (22.7)32 (14.0)39 (14.2)34 (17.8)0.113 Osteoporosis533 (5.2)341 (5.8)155 (4.1)16 (7.0)10 (3.6)11 (5.8)0.077 Recurrent urinary tract infection141 (1.4)92 (1.6)41 (1.1)1 (0.4)4 (1.5)3 (1.6)0.055 Diabetic foot ulcer334 (3.2)200 (3.4)121 (3.2)0 (0.0)7 (2.6)6 (3.1)0.116 Amputation120 (1.2)49 (0.8)68 (1.8)0 (0.0)3 (1.1)0 (0.0)0.111Service utilization Number of eligible encounters30 74016 74012 062629811498 Years in the cohort,* median (Q1, Q3)5.0 (3.9, 5.0)5.0 (3.9, 5.0)5.0 (4.0, 5.0)5.0 (3.7, 5.0)5.0 (3.5, 5.0)5.0 (3.4, 5.0)0.094 Number of eligible encounters per person year, median (Q1, Q3)0.6 (0.3, 1.0)0.6 (0.3, 1.0)0.6 (0.3, 1.0)0.5 (0.2, 1.0)0.6 (0.3, 1.1)0.5 (0.3, 0.9)0.063Medications at baseline Insulin4345 (42.0)2173 (37.3)1892 (49.5)75 (32.9)129 (47.1)76 (39.8)0.177 Antihyperglycemic medications7071 (68.4)4132 (70.9)2453 (64.2)184 (80.7)187 (68.2)115 (60.2)0.212*Number of person years is calculated as follows: (1) the start date is the patient’s first active date in the registry if it is after January 1, 2018, otherwise January 1, 2018; (2) the end time is the inactive date in the registry if it is before December 31, 2022, otherwise December 31, 2022.BMI, body mass index; CKD, chronic kidney disease; NH, non-Hispanic.Among 10 345 adults in the analysis cohort, 53% were male, with a median (25th, 75th) age of 67 years (58, 74). The racial and ethnic composition included 56% non-Hispanic White, 37% non-Hispanic Black, 3% Hispanic, 2% Asian, and 2% from other race and ethnicity groups. Less than 3% of the cohort were uninsured, 38% having commercial insurance, 29% with Medicare Advantage, 20% with traditional Medicare, and 8% having Medicaid. There were differences in insurance payer status by race and ethnicity. Compared with White patients, Black patients were younger, more likely to be female, more likely to be insured by Medicaid, and less likely to have commercial insurance. Black patients were also more likely to reside in neighborhoods with higher ADI values.Among the study population, the median (25th, 75th) A1c was 8.7 (8.3, 9.8), the median body mass index (BMI) was 31.5 kg/m2 (27.4, 36.6), and the median eGFR was 60 mL/min/1.73 m2 (48, 71). The median microalbumin/creatinine ratio was 30 (17, 148). Non-Hispanic Black and Hispanic patients had higher A1c levels compared with other race and ethnicity groups at baseline. Asian patients had lower BMI values compared with other races and ethnic groups. The eGFRs were similar across racial and ethnic groups.At baseline, the prevalence of clinical indications for prescribing SGLT2i and/or GLP-1RA was as follows: 28.4% had ischemic heart disease, 12.9% had a history of stroke, 51.4% had diabetic renal disease, 19.8% had congestive heart failure, and 26.9% had peripheral vascular disease. At baseline, 42% of patients were prescribed insulin, with higher use among non-Hispanic Black (49.5%) and Hispanic (47.1%) patients compared with White (37.3%) and Asian (32.9%) patients.Sodium-Glucose Cotransporter-2 Inhibitors There were 8219 patients with 23 453 encounters that were considered eligible for an SGLT2i prescription. Rates of new SGLT2i prescriptions per encounter significantly increased from 0.9% in 2018 to 4.3% in 2022. There were racial differences in the overall rates of new prescriptions, with Asian patients having the highest prescribed rate (3.4%), followed by non-Hispanic White patients (3.0%), Hispanics (2.3%), and non-Hispanic Black populations having the lowest rate at 1.7%. In unadjusted models and models adjusted for sociodemographic, clinical, and service utilization factors, there were significant racial disparities in new SGLT2i prescriptions. Non-Hispanic Black patients were less likely to receive new SGLT2i prescriptions compared with White patients in all models (unadjusted RR 0.61, 95% CI 0.48 to 0.78; sociodemographic RR 0.56, 95% CI 0.41 to 0.77; sociodemographic and clinical RR 0.58, 95% CI 0.41 to 0.81; fully adjusted RR 0.59, 95% CI 0.42 to 0.82). Results from the unadjusted and adjusted models are shown in table 2.Table 2Unadjusted and adjusted relative risks for new SGLT2i or GLP-1RA prescription on the encounter levelSample size: patients/encountersOverall new Rx rateUnadjusted model relative risk (95% CI)Sociodemographic-adjusted* modelrelative risk (95% CI)Clinical-adjusted† modelrelative risk (95% CI)Service-adjusted‡ modelrelative risk (95% CI)SGLT2i Race and ethnicity  NH White4292/112182.99%Reference  NH Black2778/81821.65%0.61 (0.48 to 0.78)***0.56 (0.41 to 0.77)***0.58 (0.41 to 0.81)**0.59 (0.42 to 0.82)**  NH Asian158/3813.41%1.10 (0.56 to 2.17)1.07 (0.50 to 2.27)1.14 (0.54 to 2.44)1.14 (0.54 to 2.41)  Hispanics203/5212.30%0.98 (0.53 to 1.80)0.67 (0.35 to 1.26)0.66 (0.35 to 1.26)0.65 (0.34 to 1.23)  Other140/3041.32%0.53 (0.19 to 1.44)0.42 (0.16 to 1.13)0.42 (0.16 to 1.14)0.43 (0.16 to 1.14) Year  20182769/42250.92%Reference  20193228/52031.81%2.24 (1.66 to 3.02)***1.96 (1.33 to 2.88)***1.95 (1.32 to 2.87)***1.95 (1.32 to 2.88)***  20202684/39232.68%3.09 (2.22 to 4.29)***2.92 (1.94 to 4.41)***2.89 (1.92 to 4.35)***2.86 (1.90 to 4.31)***  20212723/40853.04%3.60 (2.59 to 5.01)***3.28 (2.17 to 4.96)***3.28 (2.17 to 4.95)***3.47 (2.30 to 5.24)***  20222205/31704.32%5.26 (3.77 to 7.33)***4.70 (3.13 to 7.05)***4.67 (3.11 to 7.03)***4.95 (3.30 to 7.41)***GLP-1RA Race and ethnicity  NH White4512/113804.02%Reference  NH Black2880/76232.47%0.68 (0.56 to 0.83)***0.55 (0.44 to 0.69)***0.58 (0.46 to 0.73)***0.62 (0.49 to 0.79)***  NH Asian176/4462.02%0.50 (0.21 to 1.14)0.50 (0.21 to 1.23)0.66 (0.27 to 1.61)0.66 (0.27 to 1.59)  Hispanics200/4882.25%0.62 (0.33 to 1.17)0.46 (0.24 to 0.88)*0.49 (0.26 to 0.94)*0.48 (0.25 to 0.92)*  Other134/3193.45%0.78 (0.35 to 1.74)0.80 (0.30 to 2.15)0.78 (0.29 to 2.07)0.81 (0.30 to 2.18) Year  20182847/42941.21%Reference  20193286/51692.61%2.20 (1.69 to 2.88)***2.16 (1.58 to 2.96)***2.22 (1.62 to 3.03)***2.26 (1.65 to 3.11)***  20202725/39113.35%2.96 (2.24 to 3.92)***2.79 (2.00 to 3.88)***2.83 (2.03 to 3.95)***2.89 (2.07 to 4.02)***  20212674/38545.24%4.64 (3.51 to 6.12)***4.43 (3.19 to 6.16)***4.55 (3.27 to 6.32)***5.02 (3.60 to 6.98)***  20222172/30285.15%4.75 (3.57 to 6.32)***4.48 (3.18 to 6.32)***4.77 (3.38 to 6.72)***5.37 (3.81 to 7.57)***The GLP-1RA models were fit on 20 256 unique encounters (complete cases across all variables). The SGLT2i models were fit on 20 606 unique encounters (complete cases for all variables).*** p <.001, ** p <.01, * p <.05*Adjusted for age, sex, insurance status, area deprivation index.†Additionally adjusted for A1c level at the encounter, microalbumin/creatinine ratio, glomerular filtration rate, comorbidities, and body mass index.‡Additionally adjusted for provider type, whether the clinic has an embedded pharmacist, and service utilization history (number of out-of-control eligible A1c encounters in the past year, number of emergency department visits, outpatient visits, primary care physicians visits, endocrinologist visits, and hospitalization in the past 6 months, respectively).GLP-1RA, glucagon-like peptide-1 receptor agonists; NH, non-Hispanic; SGLT2i, sodium-glucose cotransporter-2 inhibitors.Glucagon-Like Peptide-1 Receptor Agonists There were 8634 patients with 23 232 encounters that were considered eligible for a GLP-1RA prescription. Rates of new GLP-1RA prescriptions per encounter significantly increased from 1.2% in 2018 to 5.2% in 2022. As with SGLT2is, racial differences were observed in rates of new prescriptions, with non-Hispanic White patients having the highest rate (4.0%), followed by other race and ethnicity groups (3.5%), non-Hispanic Black (2.5%), Hispanic (2.3%), and Asian patients (2.0%). In both unadjusted and adjusted models, racial disparities were present. Non-Hispanic Black patients were less likely to have received a new GLP-1RA prescription compared with White patients in all models (unadjusted RR 0.68, 95% CI 0.56 to 0.83; sociodemographic RR 0.55, 95% CI 0.44 to 0.69; sociodemographic and clinical RR 0.58, 95% CI 0.46 to 0.73; fully adjusted RR 0.62, 95% CI 0.49 to 0.79). Hispanic patients were also less likely to receive a new GLP-1RA prescription in all adjusted models (sociodemographic RR 0.46, 95% CI 0.24 to 0.88; sociodemographic and clinical RR 0.49, 95% CI 0.26 to 0.94; fully adjusted RR 0.48, 95% CI 0.25 to 0.92). Results from the unadjusted and adjusted models are shown in table 2.Insurance status and prescribing In unadjusted analyses using commercial insurance as the reference group, patients with traditional Medicare were less likely to receive new SGLT2i and GLP-1RA prescriptions, and uninsured patients were less likely to receive new SGLT2i prescriptions. Medicaid and Medicare Advantage coverage were not significantly associated with prescribing for either medication class. Adjustment for insurance status and other sociodemographic and clinical factors did not meaningfully attenuate the observed racial and ethnic disparities in prescribing. These findings suggest that differences in payer mix alone do not fully explain the observed disparities.Time trends of racial disparities in prescribing The estimated probability of new GLP-1RA and SGLT2i prescriptions increased for both non-Hispanic White and non-Hispanic Black patients from 2018 to 2022. Despite an overall increase in new prescription rates, racial disparities in new GLP-1RA prescriptions persisted among non-Hispanic Black patients, with a consistently lower probability observed throughout the study period, even in the fully adjusted model. In contrast, disparities in SGLT2i prescribing narrowed over time and were no longer statistically significant in later years. Plots of predicted prescription rates are shown in figure 2.Figure 2Model-based time trends in probability of GLP-1RA and SGLT2i prescriptions: by race/ethnicity. Unadjusted probability estimates are shown in panels (A) and (C) for GLP-1RA and SGLT2i prescriptions, respectively. Adjusted probabilities are estimated at the mean value of continuous variables and mode of categorical in panels (B) and (D) for GLP-1RA and SGLT2i prescriptions, respectively. Shaded bands indicate 95% prediction intervals around the probabilities. GLP-1RA, glucagon-like peptide-1 receptor agonist; SGLT2i, sodium-glucose cotransporter-2 inhibitor.Sensitivity analyses Results were unchanged when analyses were restricted to encounters with HbA1c values obtained prior to the encounter, with no meaningful differences in effect estimates or statistical significance. Applying time-specific eGFR exclusion criteria resulted in the exclusion of 5077 encounters (16.5% of the original analytic sample). Associations between race and ethnicity and new SGLT2i and GLP-1RA prescribing were unchanged in direction and statistical significance compared with the primary analysis, with only minimal differences in effect estimates ( online supplemental tables 3 and 4).Conclusion We observed that while prescribing rates of SGLT2i and GLP-1RA have increased over time, substantial therapeutic inertia persisted throughout the study period among patients with type 2 diabetes, above-target HbA1c, and compelling indications for use (eg, ischemic heart disease, heart failure, and high cardiovascular risk) during a period of evolving evidence for kidney-related benefits. Racial disparities in new prescription rates of both medication classes were also present, with non-Hispanic Black and Hispanic populations having disproportionately lower rates compared with non-Hispanic White populations. This racial prescribing gap persisted for GLP-1RA but was attenuated in later years for SGLT2i. Our analysis also revealed a signal toward lower prescribing rates of GLP-1RA (but not SGLT2i) among Asian populations, though this finding did not reach statistical significance, likely due to the smaller sample size of Asian patients in our cohort. These gaps represent missed opportunities to reduce cardiovascular and renal complications among high-risk populations and may perpetuate existing disparities in diabetes-related outcomes.Our work builds on prior research documenting underutilization and inequities in access to these guideline-recommended therapies, while offering several novel contributions.25–29 34–36 We focused specifically on patients with above-target HbA1c and compelling indications, allowing us to isolate prescribing patterns that reflect therapeutic inertia rather than patient ineligibility. This approach contrasts with prior studies that primarily assess cumulative or ‘ever-use’ of these agents, which may conflate delayed adoption with clinical inaction. We analyzed prescribing trends over a 5-year period marked by evolving clinical guidelines, during which major randomized controlled trials demonstrated significant cardiorenal and mortality benefits of SGLT2i and GLP-1RA drugs, leading the ADA to shift from recommending these drugs as second-line agents (following metformin as first-line therapy) to endorsing their use regardless of metformin use for patients with high cardiometabolic risk. While prescribing increased in parallel with these changes, disparities in new GLP-1RA prescriptions among Black patients remained unchanged, raising concerns about the equity of treatment adoption.Prior work has cited potential barriers to guideline-directed adoption of these agents to include differential access to quality diabetes care, including specialists who are familiar with the benefits of these medication classes, insurance and pharmacy coverage restrictions, patients’ mistrust of medical systems or fear of medication side effects, lack of shared decision-making, poor communication, and provider bias that certain groups of patients may be less likely to be adherent to treatment with an expensive agent.25 35 37 While we adjusted for whether patients were seen by a primary care provider or endocrinologist, differences in prescribing persisted, highlighting that efforts to improve uptake must address both system-wide therapeutic inertia and inequities embedded within clinical decision-making.This study has several strengths. The large, diverse sample size and multiyear scope allowed us to detect trends and disparities with precision. The use of a nested analysis approach and adjustment for absolute and relative contraindications allows for an accurate representation of real-world clinical practices, providing insight into how therapies are being prescribed and the influence of patient, provider, and health system variables. This level of granularity strengthens the relevance of our findings for informing interventions aimed at improving guideline-directed prescribing.While this study offers critical insights, there are limitations that warrant consideration. As with any EHR-based study, we were unable to capture certain factors that may influence prescribing decisions, including patient treatment preferences, quality of provider communication, shared decision-making processes, provider recommendations, or perceptions of medication cost and risk. These unmeasured interpersonal dynamics may vary across racial/ethnic groups and could contribute to inequities in treatment. Additionally, other potential clinical considerations, such as frailty, cancer history, or clinician concerns about weight loss, were not used as exclusion criteria, as these factors are not formal contraindications under current diabetes guidelines and are not reliably captured in structured EHR data. Misclassification within the EHR is also possible; however, such errors would be expected to affect all racial and ethnic groups similarly. In addition, prescriptions written outside the DUHS may not have been fully captured. To mitigate this, we restricted the analytic cohort to patients with HbA1c testing within 10 days of an ambulatory encounter, increasing the likelihood of identifying patients receiving routine diabetes care within the health system. Guideline uptake and clinical acceptance of kidney-related indications for GLP-1RAs were evolving during the study period, which may have contributed to lower initiation rates earlier in the analytic window; however, this temporal uncertainty does not explain the persistent racial and ethnic disparities observed across the study period. Finally, our eligibility criteria for SGLT2i initiation excluded patients with a history of amputation or osteoporosis as a conservative approach, although these risks are most strongly associated with canagliflozin rather than the SGLT2i class as a whole. This approach may have led to a slight underestimation of overall SGLT2i eligibility but did not differentially affect racial and ethnic comparisons.Identifying and addressing the drivers of therapeutic inertia in prescribing SGLT2i and GLP-1RA is critical to improving diabetes management and outcomes. To better understand barriers to uptake of these medication classes, future work will incorporate longitudinal quantitative and qualitative methods to examine the role of cost, patient–clinician relationships, treatment preferences, and decision-making processes. Implementation efforts will include developing clinical decision support tools to mitigate provider bias, expanding patient access to these therapies through population health strategies, and advocacy work to improve affordability. These actions are essential for ensuring access to these life-saving therapies and improving outcomes for all patients with diabetes.", "title": "Low uptake and disparities in therapeutic inertia of cardiorenal protective diabetes medications for patients with type 2 diabetes and above-target hemoglobin A1c", "uid": "23a2da20-4f16-5443-adb0-933a2832498d" }