This observational, cross-sectional study analyzed clinical and demographic data retrospectively extracted from electronic medical records (EMR) of adult patients attending a community-based primary care clinic (Caridad Center, Palm Beach, Florida, USA) between 2022 and 2024. In addition, to contextualize the study period, key descriptive indicators were summarized by calendar year (2022, 2023, and 2024).

Eligible participants were adults (≥18 years) who received at least two medical visits at the clinic during the study period. During the study period, 8,862 adult patients were identified in the EMR. Of these, 1,397 met the criterion of having a confirmed diagnosis of T2DM, resulting in a final analytical sample of not-treated (n=564) and treated (n=833) (Fig. 1). For this analysis, only patients with a confirmed diagnosis of T2DM were included (n=1,397). Diagnosis was confirmed based on laboratory evidence (for example, HbA1c ≥6.5% or fasting plasma glucose ≥126 mg/dl) consistent with ADA criteria (3,17-19). Patients with incomplete records, defined as missing essential clinical or biochemical data, were excluded. Data quality was independently validated by two clinically trained researchers, and records with inconsistent or implausible values were systematically removed to ensure accuracy and reliability. Importantly, clinical staff at the clinic, who were not part of the research team, conducted the initial de-identification of EMR prior to sharing data for analysis, thereby ensuring blinding of the investigators.

Clinical, demographic and biochemical data were systematically collected from the EMR. Demographic and socioeconomic variables included age, sex, education level, employment status, household composition and lifestyle behaviors (alcohol and tobacco use). Clinical variables focused on comorbidities, including hypertension, obesity, chronic kidney disease (CKD), hyperlipidemia, cardiovascular disease, depression/anxiety, and liver disease, which were used as adjustment covariates in regression models (20,21). Biochemical markers included glycated hemoglobin (HbA1c) to evaluate glycemic control, and renal function parameters, estimated glomerular filtration rate (eGFR) and UACR (22). Data on duration of diabetes and medication adherence were not consistently available and therefore not analyzed. Healthcare utilization during the study period (2022-2024) was recorded, including primary care visits, specialist consultations, diabetes education sessions, mental health visits, and case management encounters. These measures were primarily used as covariates for adjustment in multivariable analyses.

Medication data included prescribed glucose-lowering agents: Insulin, biguanides (metformin), sulfonylureas, sodium-glucose cotransporter-2 (SGLT2) inhibitors, glucagon-like peptide-1 (GLP-1) receptor agonists (RAs), and dipeptidyl peptidase-4 (DPP-4) inhibitors. Mineralocorticoid receptor antagonists (MRAs) were also captured as a medication class documented in the EMR and reported descriptively. Patients were categorized by treatment regimen complexity into monotherapy, dual therapy, triple therapy, or regimens with ≥4 medications, to assess treatment intensity (23-26). Fixed-dose combinations were counted according to the number of active agents contained. Medication doses and adherence data were not available.

Medication exposure was defined cross-sectionally based on active prescriptions recorded during the full 2022-2024 period; therefore, year-stratified medication patterns are not emphasized because prescriptions may span multiple calendar years and do not capture initiation, discontinuation, or sequencing.

Information on treatment duration, sequencing of therapies, prior medication changes, dosage adjustments, and medication adherence was not consistently available in the EMR and therefore could not be analyzed. As a result, medication classes were treated as static exposure variables, and analyses stratified by medication type reflect cross-sectional associations rather than longitudinal treatment effects. Therefore, medication-related findings should not be interpreted as evidence of therapeutic effectiveness.

The primary outcome was optimal glycemic control, defined as an HbA1c ≤7% (27). For patients with more than one HbA1c measurement during the study period, the lowest recorded HbA1c value was selected to represent their best-achieved glycemic status. HbA1c measurements per patient during the study period were also counted. This approach was chosen to descriptively capture the best glycemic value recorded during the study period under routine clinical care. Secondary outcomes included: i) Renal function parameters: Best-achieved eGFR and UACR; ii) treatment regimen comparisons: Glycemic control rates stratified by regimen complexity (monotherapy, dual, triple, ≥4 agents).

Covariates included age, sex, tobacco and alcohol use, employment status, comorbidities (obesity, hyperlipidemia, hypertension and CKD), healthcare utilization variables, and use of major medication classes (insulin, sulfonylureas, biguanides, GLP-1 RAs and SGLT2 inhibitors).

Descriptive statistics summarized demographic, clinical, biochemical and treatment characteristics. Comparative analyses were conducted using independent t-tests for continuous variables and chi-square tests for categorical variables. Pairwise comparisons of glycemic control rates across treatment regimens were performed using Fisher's exact test (28). A multivariable logistic regression model was used to identify independent factors associated with optimal glycemic control (HbA1c ≤7%). Covariates included age, sex, tobacco and alcohol use, employment status, comorbidities (obesity, hyperlipidemia, hypertension and CKD), healthcare utilization variables, and use of major medication classes (insulin, sulfonylureas, biguanides, GLP-1 RAs and SGLT2 inhibitors). Results were reported as adjusted odds ratios (aOR) with 95% confidence intervals (CI). Variable selection followed a backward stepwise approach (entry criterion P≤0.05; removal criterion P≥0.10). Statistically significant difference was set at P<0.05 (two-tailed). Multivariable logistic regression models used a complete-case approach (listwise deletion), and no imputation was performed. Analyses were conducted using SPSS software (version 26; IBM Corp.) (29).

The present study was approved by the Florida International University Institutional Review Board (approval no. IRB-24-0488; Miami, USA). The requirement for individual informed consent was waived due to the retrospective nature of the analysis. All procedures complied with ethical standards for research involving human subjects and ensured strict confidentiality of patient data. The present study was not preregistered; analyses should be interpreted as descriptive and exploratory.

The study population included a total of 1,397 participants with type 2 diabetes. Demographic, clinical and healthcare utilization characteristics among patients with type 2 diabetes (T2DM) according to their use of glucose-lowering medications are presented in Table I. Patients receiving pharmacological treatment were significantly older and had a higher prevalence of comorbidities, including CKD, hypertension, hyperlipidemia and obesity, compared with non-medicated patients (all P<0.05). Furthermore, medicated patients exhibited worse glycemic control, as reflected by higher mean HbA1c levels, and lower renal function, indicated by reduced eGFR values (both P<0.001). Healthcare utilization was significantly greater among treated patients, who had more primary care appointments, specialist visits, case management encounters, mental health visits, and health education sessions. Although the proportion of females was slightly lower in the medicated group (59.5%) than in the non-medicated group (64.2%), this difference was not statistically significant (P=0.083). During 2022-2024, the number of HbA1c measurements per patient had a median of 2.0 (IQR 2), with minimum and maximum values of 1 and 13, respectively.

Medication use among the treated subgroup (n=833) of patients with T2DM is shown in Table II. Within these patients, metformin was the most prescribed medication, used by 86.4% of patients, followed by insulin (25.5%), GLP-1 RAs (24.1%) and SGLT2 inhibitors (19.3%). Sulfonylureas were prescribed to 17.4% of patients treated, while MRAs and DPP-4 inhibitors were less frequently used, at 2.2 and 0.1%, respectively. Regarding therapy regimens, just over 50% of the patients were on monotherapy (51.4%), while 28.9 and 14.1% were on dual and triple therapy, respectively. A smaller proportion (4.6%) received four or more medications.

In Table III, the proportion of patients achieving optimal glycemic control (HbA1c ≤7%) according to medication class and regimen complexity is shown. Among patients receiving monotherapy, the highest control rates were observed for SGLT2 inhibitors (90.9%) or GLP-1 Ras (85.7%), followed closely by biguanides (85.5%). Sulfonylureas showed intermediate control rates, (62.5%), while insulin monotherapy revealed notably lower rates of glycemic control (33.3%). Pairwise comparisons of glycemic control among monotherapy regimens using Fisher's exact test are demonstrated in Table IV. Higher proportions of glycemic control were observed among patients receiving SGLT2 inhibitors or GLP-1 RAs compared with those receiving insulin monotherapy; however, these differences likely reflect variation in underlying disease stage and treatment selection (P<0.001 for both comparisons). Biguanides also showed a higher proportion of glycemic control than insulin (P<0.001). Although sulfonylureas showed intermediate glycemic control rates, differences compared with other medications were not statistically significant, highlighting intermediate control rates in monotherapy.

Table V shows, for each comorbidity, the proportion of patients with (Yes) or without (No) the condition who receive monotherapy with each medication class. Among patients with CKD, a significantly higher proportion received SGLT2 inhibitors (20.5%) compared with those without CKD (2.8%) (P<0.001). GLP-1 RAs are more frequently prescribed to patients with obesity, with 13.1% of obese patients receiving this therapy compared with 1.0% of non-obese patients (P<0.001). Biguanide use is dominant among patients without CKD (85.7%) and less common in those with CKD (38.5%) (P<0.001); similarly, biguanides are more frequent in patients without hypertension than in those with it. Sulfonylureas show no significant differences in prescription rates between patients with or without the evaluated comorbidities. Insulin monotherapy is characteristically prescribed to patients with more complex clinical states, such as CKD and possibly lower metabolic reserve, as suggested by the lower prevalence of obesity among insulin users (1.4%) compared with non-users (8.0%) (P=0.004). This is evidenced by the markedly higher proportion of patients with CKD receiving insulin (25.6%) vs. those without CKD (3.8%) (P<0.001). Additionally, insulin use is significantly higher among unemployed patients (8.6%) compared with employed ones (4.0%) (P=0.047), which may reflect differences in clinical complexity, health care access, or other unmeasured factors. These data suggest that prescribing patterns reflect selective use of SGLT2 inhibitors and GLP-1 RAs in specific clinical phenotypes (such as CKD and obesity) rather than uniform superiority across disease stages, while insulin monotherapy is commonly observed among patients with more advanced clinical profiles of advanced diabetes, associated with more severe comorbidities and complex patient profiles. Notably, patients with regimens that include the use of insulin, either as monotherapy or in combination with other medications, are those with the lowest proportion of glycemic control (Table III).

HbA1c, eGFR and UACR among patients treated with different monotherapy regimens are compared in Table VI. Significant differences in HbA1c were observed, with the lowest mean values found in patients treated with SGLT2 inhibitors (5.95%), GLP-1 RAs (6.28%), and biguanides (6.38%). Patients treated with insulin (7.89%) and sulfonylureas (7.92%) had significantly higher mean HbA1c levels, indicating lower glycemic control (P<0.001 for comparisons). By contrast, renal function parameters (eGFR and UACR) appeared indicative of baseline patient characteristics rather than therapeutic effects. Patients on insulin monotherapy had significantly lower eGFR (67.7 ml/min/1.73 m²) compared with those on biguanides (101.1 ml/min/1.73 m²; P<0.001) and GLP-1 RAs (102.5 ml/min/1.73 m²; P<0.001). Patients receiving SGLT2 inhibitors showed the lowest eGFR (59.9 ml/min/1.73 m²), indicating targeted prescription among patients with advanced renal impairment. The UACR, reflecting renal damage, was significantly elevated in patients treated with sulfonylureas (349.0 mg/g) and insulin (148.6 mg/g), compared with notably lower values in those receiving biguanides (30.6 mg/g; P<0.001), GLP-1 RAs (47.3 mg/g), and SGLT2 inhibitors (66.6 mg/g; P=0.047). In summary, it can be stated that patients receiving insulin or sulfonylureas had higher HbA1c levels and less favorable renal parameters; however, these differences likely reflect more advanced disease stage and selective prescribing rather than medication-related effects, consistent with more advanced disease.

Glycemic control, eGFR and UACR values among patients on dual and triple diabetes medication regimens are compared in Table VII. In dual therapy, glycemic control was significantly more favorable in regimens containing GLP-1 RAs, with a mean HbA1c of 6.68%, compared with those without GLP-1 RAs (7.92%; P<0.001). Dual therapy involving insulin was associated with significantly poorer glycemic control (HbA1c, 8.51%) compared with those without insulin (7.07%; P<0.001). Dual regimens including sulfonylureas did not show significant glycemic advantages (7.41 vs. 7.58%; P=0.549). Triple therapy regimens including sulfonylureas demonstrated significantly poorer glycemic outcomes (HbA1c, 8.12%) compared with regimens without sulfonylureas (7.50%; P=0.035). Although triple therapies involving GLP-1 RAs had slightly improved glycemic control (7.55%) compared with regimens without GLP-1 RAs (7.97%), this difference was not statistically significant (P=0.138). Regarding renal function parameters, patients receiving dual therapy with biguanides had significantly improved eGFR (102.0 ml/min/1.73 m² vs. 79.2 ml/min/1.73 m² without biguanides; P<0.001). Similarly, dual therapy involving GLP-1 RAs demonstrated significantly improved eGFR (104.5 ml/min/1.73 m² vs. 95.6 ml/min/1.73 m² without GLP-1; P=0.018). Patients treated with dual therapies containing SGLT2 inhibitors had significantly lower mean eGFR (81.4 ml/min/1.73 m² vs. 102.9 ml/min/1.73 m²; P<0.001), reflecting their targeted use among patients with impaired renal function. The UACR was significantly lower in dual therapies with GLP-1 RAs (24.1 mg/g vs. 95.2 mg/g; P=0.023) and biguanides (31.4 mg/g vs. 267.8 mg/g; P<0.001), showing favorable renal parameter profiles in this cross-sectional comparison. In triple therapy, higher UACR values among patients on SGLT2 inhibitors (88.7 mg/g vs. 32.0 mg/g without SGLT2 inhibitors; P=0.044) likely reflect selective prescribing for those with advanced renal disease.

Factors significantly associated with optimal glycemic control (defined as HbA1c ≤7%) among patients with diabetes are presented in Table VIII. In the multivariable logistic regression model, female sex (aOR=3.246, 95% CI: 1.323-7.964; P=0.010) and frequent specialist visits (more than nine visits, aOR=8.267, 95% CI: 2.344-29.164; P=0.001) were significantly associated with better glycemic control. By contrast, insulin use (aOR=0.018, 95% CI: 0.005-0.060; P<0.001) and sulfonylurea use (aOR=0.072, 95% CI: 0.019-0.268; P<0.001) were strongly associated with poorer glycemic control. These results suggest that insulin and sulfonylureas are predominantly used among patients with more advanced disease, greater therapeutic complexity, or non-optimal therapeutic regimen, reflecting clinical challenges in achieving optimal glucose control. Meanwhile, increased specialist care and certain demographic factors (such as female sex) were important factors associated with more favorable glycemic management. Overall, these findings underscore the importance of targeted healthcare resource utilization and careful medication selection in optimizing glycemic outcomes among diabetic patients.

Of the 1,397 patients included in the analysis, a total of 4,099 HbA1c laboratory determinations were recorded during the 2022-2024 study period. The annual distribution was 1,202 measurements in 2022 (29.3%), 1,347 in 2023 (32.9%), and 1,550 in 2024 (37.8%), demonstrating a progressive increase in the absolute number of HbA1c assessments over time.

Regarding cumulative measurement frequency per patient across the entire study period, 138 patients (9.9%) had one determination, 564 (40.4%) had two, 327 (23.4%) had three, and 368 (26.3%) had four or more measurements.

Overall, these descriptive findings indicate an increase in outpatient laboratory monitoring activity throughout the study period. Although this pattern is consistent with the gradual normalization of healthcare delivery following the acute phase of the COVID-19 pandemic, no formal temporal trend analysis was conducted; therefore, this interpretation should be considered exploratory rather than causal evidence of post-pandemic recovery or the absence of temporal sampling bias. During 2022-2024, the number of HbA1c measurements per patient had a median of 2.0 (IQR 2), with minimum and maximum values of 1 and 13, respectively.

Healthcare utilization patterns varied notably according to the type of pharmacological treatment received by patients with T2DM. Those treated with insulin, SGLT2 inhibitors, or GLP-1 RAs exhibited significantly higher use of healthcare services, including primary care, specialist consultations, case management encounters and health education sessions (all P<0.001) (Table IX). Insulin users, in particular, showed the highest engagement in specialist visits (28.21±21.51) and case management services (6.61±5.45), reflecting insulin's typical use in patients with greater clinical complexity who require closer follow-up. Similarly, patients treated with SGLT2 inhibitors or GLP-1 RAs demonstrated substantial healthcare resource use, which may reflect differences in clinical complexity and prescribing patterns and the relatively recent introduction of these therapies, which may necessitate more intensive monitoring. Notably, mental health service utilization was also higher among patients treated with insulin (P=0.005) and GLP-1 RAs (P<0.001), possibly reflecting the psychosocial burden associated with complex treatment regimens or disease severity. By contrast, sulfonylureas and biguanides were associated with lower use of specialist services and case management, suggesting their preferential use in patients with less complicated disease profiles or those with stable glycemic control. Particularly, metformin users had significantly fewer specialist visits (P<0.001), consistent with their common use as first-line therapy in routine clinical practice (Table IX).

Factors independently associated with the prescription of specific diabetes medications revealed distinct clinical and healthcare engagement patterns (Table X). In the adjusted multivariable logistic regression analysis, insulin use was significantly linked to the presence of hypertension (aOR=2.94; 95% CI: 1.29-6.69; P=0.010) and attendance at diabetes education sessions (aOR=4.89; 95% CI: 2.30-10.41; P<0.001). Conversely, insulin was less likely to be prescribed in combination with sulfonylureas (aOR=0.30; 95% CI: 0.10-0.87; P=0.027), likely reflecting prescribing practices that aim to reduce therapeutic complexity and the risk of hypoglycemia. The use of SGLT2 inhibitors was strongly associated with hypertension (aOR=4.18; 95% CI: 1.53-11.47; P=0.005), suggesting this is consistent with selective prescribing patterns for patients with increased cardiovascular or renal risk. Prescription of GLP-1 RAs was significantly associated with overweight or obesity (aOR=2.55; 95% CI: 1.30-5.00; P=0.006) and participation in diabetes education programs (aOR=3.72; 95% CI: 1.84-7.52; P<0.001), consistent with their known benefits in weight management and the role of patient engagement in complex therapy adherence. By contrast, biguanides (metformin) were significantly less likely to be prescribed in patients with impaired renal function (aOR=0.34; 95% CI: 0.13-0.84; P=0.019), in line with clinical guidelines recommending caution due to the risk of lactic acidosis in this population.

These findings emphasize that prescription decisions for diabetes medications are associated not only with comorbid conditions such as hypertension, renal function and obesity but also by healthcare engagement factors such as diabetes education attendance. The observed patterns reflect a personalized approach in pharmacological management, in which medication selection varies with efficacy, safety and complexity according to each patient's clinical profile (Table X).