Practice Variation in Prescribing SGLT2-Inhibitors: A Population-Based Analysis from Alberta, Canada

Abstract

Background

Sodium glucose-cotransporter 2 inhibitors (SGLT2is) have been shown to reduce the risk of cardiorenal complications in select patient populations, yet their real-world uptake in clinical practice is limited. The objective of this study was to assess the factors influencing prescriber variation in SGLT2i use.

Methods

Using administrative data from Alberta, Canada, we conducted a population-based cohort study of adults with a new prescription for any oral antihyperglycemic medication between 2014-2021. We used multilevel logistic regression to examine how patient and prescriber characteristics were associated with SGLT2i prescription and used the median odds ratio to quantify variation at the prescriber level.

Results

Of 339,314 patients prescribed a new antihyperglycemic medication, 0.8% (n = 2852) were prescribed an SGLT2i. SGLT2i prescribing was more likely among male patients, younger patients, and those with obesity or heart failure. However, substantial variation was present in prescriber behaviour for SGLT2i, and prescriber factors had a greater influence on SGLT2i prescribing than patient-level factors (median odds ratio 3.55). Although family physicians accounted for greatest numbers of SGLT2i prescriptions overall, subspecialists, such as cardiologists, were more likely to prescribe SGLT2is (odds ratio 13.5, 95% confidence interval 8.9-20.5).

Conclusions

The rate of new SGLT2i prescriptions was low during the study period. Both patient- and prescriber-level factors are associated with SGLT2i prescription, although variation in prescribing SGLT2i medications appears to be driven primarily by prescriber factors. Family physicians are responsible for the majority of SGLT2i prescriptions and represent a key provider group in aligning SGLT2i prescribing with guideline recommendations.

Newer generations of antihyperglycemic medications, such as sodium glucose-cotransporter 2 inhibitors (SGLT2is) have demonstrated significant cardiorenal benefit in patients with type 2 diabetes. Originally developed to lower blood glucose in diabetes, cardiovascular outcome trials have shown that SGLT2is not only improved glycemia, but also reduced the risk of cardiovascular death, nonfatal myocardial infarction, nonfatal stroke, heart failure hospitalizations, and the progression of chronic kidney disease.1, 2, 3, 4, 5

Since the publication of these landmark clinical trials, the use of SGLT2is is recommended by Diabetes Canada clinical practice guidelines as being the preferred add-on therapy to metformin when additional glucose lowering is required or if specific indications exist. The Kidney Disease: Improving Global Outcomes group also recommends use of SGLT2is in patients with concomitant diabetes, and the Canadian Cardiovascular Society recommends SGLT2i use in those with heart failure with reduced ejection fraction, regardless of diabetes status.^6,7 However, the diffusion of these medications in real-world clinical practice is limited. For example, an Albertan cross-sectional study conducted in 2021 found that less than 20% of patients with diabetes and established cardiorenal disease were prescribed an SGLT2i or a glucagon-like peptide 1 (GLP-1) receptor agonist—a medications with similar evidence for cardiorenal benefit.⁸ The slow adoption rate of SGLT2is raises the concern of a missed therapeutic opportunity to improve the health of patients with diabetes and avoid preventable cardiovascular complications. In addition, understanding the prescription patterns of these medications is even more relevant in the context of expanding indications for SGLT2is, beyond antihyperglycemic management, including improving outcomes in patients with cardiovascular disease, heart failure, and chronic kidney disease.^9,10

Low rates of SGLT2i prescription may be due to the phenomenon of therapeutic inertia, which is defined as the failure to provide specific treatment indicated for a patient’s condition due to prescriber, patient, and healthcare system factors.¹¹ Prior work from the US has identified substantial variation in physician prescribing behaviours for newer antihyperglycemic medications such as SGLT2is, even after accounting for clustering of clinical practice patterns.^12,13 Therefore, the objective of this study was to better understand the prescriber and patient factors associated with SGLT2i prescriptions and determine the influence of prescriber variation on SGLT2i use within the Canadian context.

Methods

Data sources

This population-based cohort study used de-identified administrative health data from Alberta, Canada held within the Interdisciplinary Chronic Disease Collaboration data repository, which includes provincial laboratory and linked administrative health data (including patient and prescriber demographics, vital statistics, prescription drug dispensations, physician claims, hospitalizations, emergency department and ambulatory visits) of all those who were eligible for public health insurance.¹⁴ This study was approved by the Conjoint Health Research Ethics Board at the University of Calgary (REB16-1575). As this study used de-identified data, the ethics board did not require written patient consent.

Study cohort

The study cohort included all adults aged > 18 years with a new prescription for an antihyperglycemic medication between April 1, 2014 and March 30, 2021. New prescriptions were defined as at least 2 dispensations of the same antihyperglycemic medication, each for at least 30 days, and no more than 365 days apart, with no dispensations for that same medication class for 12 months prior.¹² Classes of antihyperglycemic medications were identified in the administrative datasets using anatomic therapeutic chemical classification (ATC) codes, including metformin, sulfonylureas, thiazolidinediones, insulin, meglitinides, dipeptidyl peptidase 4 inhibitors, alpha-glucosidase inhibitors, GLP-1 receptor agonists, and SGLT2is.

Covariates

Baseline patient-level covariates associated with each prescription were obtained and included age, sex, urban vs rural location, socioeconomic status (defined as neighbourhood income quintile data), and medical comorbidities defined using previously validated administrative algorithms.¹⁵

Prescriber-level covariates included the following: prescriber age, sex, years in practice, and Canadian vs non-Canadian country of medical school training; clinical workload (ie, low, medium, high based on tertile of annual outpatient consultations); physician payment model if applicable (ie, fee-for-service or salaried); and practice speciality (ie, family medicine, endocrinology, cardiology, general internal medicine, and other).

Statistical analyses

Descriptive statistics were used to report the baseline characteristics of those who were prescribed an SGLT2i compared to those who received a different antihyperglycemic prescription. Patients with simultaneous prescriptions (SGLT2i and non-SGLT2i) were categorized into the SGLT2i group. Categorical variables were described as frequencies and percentages. Continuous variables were presented as mean and standard deviation, with medians and interquartile ranges presented for non-normal distributions. Χ² tests were used to compare categorical variables between groups, and Kruskal-Wallis tests were used for continuous variables. To understand the differences between prescribers, we also explored characteristics such as age, sex, speciality, and clinical workload among high-, medium-, and low-volume prescribers of SGLT2is. High-, medium-, and low-volume prescribers were defined by their tertile of overall number of prescriptions.

Multilevel logistic regression models were used to ascertain the odds of prescribing an SGLT2i. Specifically, we created a 2-level model including patient and prescriber characteristics as fixed effects and a random intercept to account for patients clustered within prescribers and to measure prescriber variation. We estimated variation using the intraclass correlation coefficient, which described unexplained variation between prescribers after accounting for known covariates included in the model as described above.¹⁶

To better understand the influence of prescriber factors on SGLT2i prescribing, we estimated the median odds ratios (MORs), which represent the increase in odds of receiving an SGLT2i in comparing 2 patients with identical characteristics treated by different prescribers—one with a greater and one with a lesser tendency to prescribe an SGLT2i.¹⁷ The MOR quantifies the magnitude of the variation between prescribers and is directly comparable to the odds ratios (ORs) of other covariates in the model.¹⁸ In this study, the MOR reflects how much a patient’s odds of receiving an SGLT2i prescription change if he or she was seen by a prescriber more likely to prescribe an SGLT2i, vs one less likely to do so. Statistical analyses were performed with Stata/IC, version 14.2 (StataCorp, College Station, TX).

Results

Study cohort

Of 339,314 patients with a new antihyperglycemic prescription, SGLT2is accounted for only 0.8% of prescriptions (n = 2852; Fig. 1). Among the 2852 patients prescribed SGLT2i medications, the mean age was 55.2 ± 15.1 years, and 60.3% (n = 1720) were male (Table 1). The most common patient comorbidities included diabetes (65.7%; n = 1875), hypertension (54.4%; n = 1552), and obesity (21.9%; n = 625).

Figure 1.

Flow chart of inclusion criteria for new prescription fills of diabetes medications. SGLT2i, sodium glucose cotransporter 2 inhibitor; GLP1RA, glucagon-like peptide 1 receptor antagonist. ∗New diabetes drug fill defined as ≥ 2 fills for any diabetes medication at least 30 days apart but no more than 365 days apart, with no fills in the year prior.

Table 1.

Patient characteristics

Characteristics	Antihyperglycemic medications (other than SGLT2is) (n = 336,462)	SGLT2i (n = 2852)
Age, y, mean (SD)	57.5 (15.1)	55.2 (15.1)
Age, y
< 65	67.3 (226,541)	78.5 (2238)
≥ 65	32.7 (109,921)	21.5 (614)
Female sex	47.2 (158,9258)	39.7 (1132)
Comorbidities
Heart failure	6.8 (22,867)	7.8 (221)
Chronic kidney disease	10.7 (35,820)	8.4 (240)
Myocardial infarction	4.6 (15,593)	5.4 (155)
Stroke	8.6 (29,000)	5.5 (158)
Peripheral arterial disease	2.0 (6713)	1.4 (39)
Hypertension	62.4 (209,853)	54.4 (1,552)
Thyroid	12.9 (43,525)	10.8 (309)
Pulmonary	15.0 (50,307)	11.5 (328)
Obesity	20.4 (68,756)	21.9 (625)
Health zone
1	33.7 (113,319)	37.7 (1,074)
2	11.5 (38,805)	14.6 (417)
3	34.2 (115,162)	30.5 (870)
4	12.6 (42,354)	8.2 (233)
5	7.6 (25,454)	7.8 (222)
Missing	0.4 (1368)	1.3 (36)
Geography
Urban	76.9 (258,735)	80.2 (2,289)
Rural	22.7 (76,359)	18.6 (529)
Missing	0.4 (1,368)	1.3 (36)
Neighbourhood income quintile
Lowest	26.1 (87,703)	22.4 (640)
2nd	22.2 (74,593)	21.0 (599)
3rd	18.0 (60,419)	18.8 (536)
4th	17.2 (58,016)	18.2 (518)
Highest	15.9 (53,469)	18.0 (512)
Unknown/ Missing	0.7 (1262)	1.8 (47)
Medication classes
Metformin	72.7 (244,753)	18.9 (540)
Sulfonylureas	11.9 (39,929)	2.9 (83)
Thiazolidinediones	1.6 (5227)	0.6 (18)
Insulin	15.0 (50,447)	2.0 (57)
Meglitinides	2.7 (8902)	0.2 (5)
DPP-4	4.4 (14,895)	0.0 (0)
Alpha-glucosidase	0.2 (669)	0.1 (2)
GLP-1 receptor agonist	6.0 (20,290)	1.6 (45)
SGLT2i	0.0 (0)	100.0 (2,852)

Values are % (n), unless otherwise indicated.

DPP-4, dipeptidyl peptidase-4 inhibitor; GLP-1, glucagon-like peptide-1; SD, standard deviation; SGLT2i, sodium glucose cotransporter-2 inhibitor.

Patient factors associated with SGLT2i prescriptions

Patients were less likely to be prescribed SGLT2i medications if they were female (OR 0.70, 95% confidence interval [CI] 0.64-0.77) and if they were over age 65 years (OR 0.62, 95% CI 0.56-0.70; Table 2). Those from the highest-income neighbourhoods were more likely to be prescribed an SGLT2i, vs those from the lowest-income neighbourhoods (OR 1.31, 95% CI 1.15-1.51). Patients with obesity and heart failure were more likely to be prescribed an SGLT2i than were those without these conditions (OR 1.25, 95% CI 1.13-1.39 and OR 1.25, 95% CI 1.04-1.50, respectively).

Table 2.

Association between patient and physician characteristics and sodium glucose cotransporter 2 inhibitor prescriptions: logistic regression results

Patient characteristics		Odds ratio (95% CI)	P
Patient sex (female)
Female vs male		0.70 (0.64–0.77)	< 0.01
Patient age, y (age 65 y)
≥ 65 vs < 65		0.62 (0.56–0.70)	< 0.01
Patient household income
2nd vs 1st (lowest quintile)		1.10 (0.97–1.25)	0.14
3rd vs 1st (lowest quintile)		1.17 (1.03–1.34)	0.02
4th vs 1st (lowest quintile)		1.18 (1.03–1.35)	0.02
Highest vs 1st (lowest quintile)		1.31 (1.15–1.51)	< 0.01
Patient comorbidity
Heart failure		1.25 (1.04–1.50)	0.02
Chronic kidney disease		0.92 (0.78–1.06)	0.24
Myocardial infarction		1.10 (0.90–1.35)	0.34
Stroke		0.72 (0.60–0.87)	< 0.01
Peripheral arterial disease		0.87 (0.60–1.25)	0.44
Hypertension		0.84 (0.76–0.92)	< 0.01
Thyroid disease		0.97 (0.85–1.11)	0.70
Pulmonary disease		0.88 (1.13–1.39)	< 0.01
Obesity		1.25 (1.13–1.39)	< 0.01
Physician characteristics
Physician sex
Female vs male		0.74 (0.63–0.87)	< 0.01
Physician age, y
41–60 vs ≤ 40		0.92 (0.69–1.25)	0.62
61–80 vs ≤ 40		0.76 (0.53–1.08)	0.12
> 81 vs ≤ 40		0.44 (0.27–0.71)	< 0.01
Specialty
General internal vs family medicine		1.64 (1.19–2.25)	< 0.01
Endocrinology vs family medicine		1.29 (0.56–3.01)	0.55
Cardiology vs family medicine		13.5 (8.86–20.5)	< 0.01
Other vs family medicine		1.09 (0.74–1.64)	0.65
Years in practice
15–30 vs < 15		0.77 (0.57–1.03)	0.08
> 30 vs < 15		0.68 (0.48–1.04)	0.07
Country of training
High/upper middle-income country with similar training system to Canada vs Canadian graduate		1.33 (1.09–1.62)	< 0.01
Other high/upper middle-income country vs Canadian graduate		1.32 (1.05–1.66)	0.02
Lower middle/low-income country vs Canadian graduate		1.55 (1.26–1.91)	< 0.01
Physician zone (zone 1 as comparator)
2		1.02 (0.81–1.30)	0.85
3		0.69 (0.5–0.82)	< 0.01
4		0.60 (0.47–0.76)	< 0.01
5		0.95 (0.72–1.25)	0.71
Clinical workload
Percentage of outpatient consultation		1.01 (1.01–1.02)	< 0.001
Variance statistics
VPC or ICC—physician	0.349
MOR	3.551

CI, confidence interval; CKD, chronic kidney disease; ICC, intraclass correlation; MI, myocardial infarction; MOR, median odds ratio; PAD, peripheral arterial disease; VPC, variance partition coefficient.

Prescriber factors associated with SGLT2i prescriptions

Most prescribers (98.4%) were physicians (n = 1200). Prescriber factors associated with an increased odds of SGLT2i prescription included subspecialty practice. vs family medicine, specifically cardiology (OR 13.5, 95% CI 8.86-20.5) and internal medicine (OR 1.64, 95% CI 1.19-2.25). Other physician factors associated with increased SGLT2i prescribing included fewer years of practice, younger physician age, graduation from non-Canadian medical school, regardless of region of training programs (ie, low, lower-middle, upper-middle, or high-income countries), and male sex (Table 2).

Variation in prescriber choices

In the unadjusted multilevel modelling, 39% of variation in SGLT2i prescribing was attributable to prescriber-level characteristics (Table 2; Supplemental Table S1). After including known patient and prescriber covariates, 35% of variation at the prescriber level remained unexplained. The variation in SGLT2i prescribing between prescribers (MOR = 3.55) was of greater relevance than the impact of all patient- and prescriber-level variables, except for cardiology specialty, for understanding variation in the odds of receiving an SGLT2i prescription (Table 2).

To further understand prescribing differences among prescribers, we explored characteristics of high-, medium-, and low-volume prescribers based on tertiles of prescriptions per provider. Prescriptions for SGLT2is were provided by 1220 prescribers. The mean prescriber age of those who prescribed SGLT2is was 44.9 ± 11.2 years, and 30.7% (n = 374) were female. The majority of SGLT2i prescriptions were provided by family physicians (78.4%; n = 956), and these prescriptions were more common among fee-for-service providers (93.8%; n = 1144) with high-volume clinical practices (Table 3).

Table 3.

Characteristics of high, medium- and low-volume prescribers of sodium glucose cotransporter 2 inhibitor medications

Physician characteristics	High-volume prescriber	Medium-volume prescriber	Low-volume prescriber
N	377	265	578
Age, y, mean (SD)	45.8 (9.9)	43.4 (10.7)	45.0 (11.7)
Female sex	28.4 (107)	27.9 (74)	33.4 (193)
Specialty
Family medicine	90.2 (340)	86.0 (228)	67.1 (388)
General internal	4.5 (17)	5.7 (15)	4.2 (24)
Endocrinology	0.3 (1)	0.0 (0)	1.0 (6)
Cardiology	0.5 (2)	3.8 (10)	5.9 (34)
Other	0.0 (0)	0.8 (2)	7.2 (42)
Years in practice	4.5 (17)	3.8 (10)	14.5 (84)
< 15
15–30	31.8 (120)	48.3 (128)	43.1 (249)
> 30	52.0 (196)	35.9 (95)	38.9 (225)
Missing	16.2 (61)	15.5 (41)	18.0 (104)
Country of training	0.0 (0)	0.4 (1)	0.0 (0)
Canada
High/upper middle-income countries with similar medical training system to Canada	37.1 (140)	39.3 (104)	52.8 (305)
Other high/upper middle-income country	28.9 (109)	24.9 (66)	14.9 (86)
Lower middle/low-income country	11.4 (43)	13.6 (36)	12.1 (70)
Unknown	22.6 (85)	21.9 (58)	20.1 (116)
Missing	0.0 (0)	0.4 (1)	0.2 (1)
Physician payment model
Fee-for-service	97.6 (368)	95.5 (253)	90.5 (523)
Salary	2.4 (9)	4.5 (12)	9.5 (55)
Clinical workload, %
< 27	1.6 (6)	5.7 (15)	16.4 (95)
27–47	17.0 (64)	27.6 (73)	29.1 (168)
> 47-	76.9 (290)	63.8 (169)	40.1 (232)

Values are % (n), unless otherwise indicated. SD, standard deviation.

Discussion

In this population-based analysis of 339,314 patients with a new antihyperglycemic prescription, SGLT2i prescription was uncommon during the study time period, accounting for ∼1% of all new antihyperglycemic prescriptions among patients with diabetes. Patient-level factors associated with higher odds of receiving an SGLT2i prescription included being younger, being male, coming from a high-income neighbourhood, and having comorbidities, such as heart failure and obesity. However, substantial variation exists in provider prescribing behaviour for SGLT2i, and these differences among providers had a greater influence on SGLT2i prescribing than did patient-level factors, including baseline comorbidities. Notably, although family physicians were the highest-volume prescribers of SGLT2is, cardiology and general internal medicine specialists were more likely to prescribe SGLT2i medication. Finally, we found that a considerable amount of unexplained variation in the type of prescription patients received remained at the prescriber level, rather than in any measured patient characteristics.

Our findings are consistent with prior research conducted in cohorts with type 2 diabetes, in which patients who are younger, male, and with higher household income are more likely to be prescribed SGLT2i by physicians.19, 20, 21, 22, 23 Sex differences could be due to prescriber hesitancy due to the adverse effect of genitourinary infections, particularly genital mycotic infections, with SGLT2is.^21,24 However, newer data on SGLT2is suggest that the risk of genitourinary infections is not distributed equally among female patients, and an approach identifying specific risk factors for genital infections, such as younger age, use of hormone-replacement therapy, and previous history of genital infections, may be indicated.²⁵ Patients with obesity also were more likely to be prescribed an SGLT2i, which may reflect a preference for weight-neutral or weight-negative diabetes medications in this population, compared to older-generation antihyperglycemic agents, which are associated with weight gain (eg, sulfonylureas, meglitinides, thiazolidinediones).²⁶

During the study period, the primary indication for SGLT2i prescription was diabetes. Specifically, Diabetes Canada clinical practice guidelines recommend SGLT2i medications as add-on therapy after metformin in patients who are at high risk for cardiorenal events and those with heart failure or chronic kidney disease. This group includes patients who are aged > 60 years and have ≥ 2 cardiovascular risk factors, such as hypertension, previous myocardial infarction, and peripheral artery disease.²⁷ Our finding of an increased odds of SGLT2i prescription in heart failure is reassuring and adheres to Diabetes Canada guidelines. However, SGLT2i prescriptions were not more likely among those with chronic kidney disease, which may be due to the delayed Health Canada approval for this indication in August 2021, which fell outside of our study’s time frame. A notable concern is that classic cardiovascular risk factors, such as hypertension and established cardiovascular disease, were not associated with increased SGLT2i prescription despite Diabetes Canada guideline indications for this in older patients with diabetes. The lower odds of SGLT2i prescriptions in older patients with diabetes and cardiac comorbidities (excluding heart failure) is likely multifactorial and may reflect both therapeutic inertia and the specific requirements for SGLT2i special authorization in the public drug coverage plans for older adults in Alberta.

Regarding therapeutic inertia, a qualitative study from Australia suggests that a possible reason for this treatment-benefit paradox with SGLT2i is that some physicians underappreciate the extent of the results of recent trials showing cardiorenal protection from these medications independent of hemoglobin A1c level.²⁸ In that study, a portion of primary care physicians also expressed hesitancy due to the number of diabetes medications now available and a preference for an endocrinologist to provide advice regarding second-line therapy for patients with diabetes.

Public drug coverage in Alberta for adults aged 65 years and older is delivered mainly through Alberta Blue Cross,²⁹ which requires special authorization; prescribers must try a first-line diabetes medication for 6 months, as well as a second-line agent such as sulfonylureas, prior to qualifying for coverage. This potentially delays initiation of cardiorenal protection in any patient with Alberta Blue Cross coverage, which includes Albertans aged ≥ 65 years. These prior authorization access barriers and patient-borne costs may be less of an issue more recently with the approval of generic dapagliflozin in 2023, which does not require special authorization approval from Alberta Blue Cross.²¹ However, the low uptake rates of SGLT2is during our study time frame (up to 2021) may reflect the barriers imposed by prior authorization forms limiting uptake of medications with cardiorenal benefit.

The prescriber factors that were most strongly associated with an increased odds of SGLT2i prescription included younger prescriber age, fewer years in clinical practice (< 15 years), higher clinical volumes, and provider speciality. These factors may suggest greater comfort in prescribing these medications, due to either proximity to recent training or greater flexibility in adopting continuing medical education updates. Cardiologists may have felt more comfortable prescribing SGLT2is, as emerging evidence highlighted cardiovascular benefits independent of diabetes, broadening the perceived scope of these agents beyond glycemic management. Maitra and colleagues have previously described that a barrier to the uptake of novel medications into clinical practice is a hesitancy to change old practice habits or routines in spite of novel information.³⁰ In addition, prior studies suggest that family physicians working in resource-limited areas often raise the issue of a lack of access to specialists as a contributor to their prescribing hesitancy, with a preference for a specialist to initiate therapy.²⁸ Resource limitations may also explain the finding of geographic variation in prescribing practices, where patients located in northern Alberta (zones 3 and 4) were more less likely to be prescribed SGLT2is. Geographic variation in guideline medication adherence within Alberta has been observed in other studies,³¹ and it may be related to variation in health resources, such as access to subspecialist care, pharmacist support, or laboratory services.^31,32

Our study identifies several opportunities to improve uptake and reduce the impact of prescriber variation of SGLT2i medications in clinical practice. Focused continuing medical education in the primary care setting may facilitate prescriber confidence in initiating SGLT2i prescriptions, particularly for indications endorsed by current clinical guidelines. In addition, embedded electronic medical record practice alerts and clinical decision support tools may help guide clinical decision-making and adherence to guidelines.³³ These interventions have been noted to be successful in other areas of medicine. For example, a study by Sheth and colleagues found that electronic medical record practice alerts improved pneumococcal vaccination rates in immunosuppressed patients, from 28% to 62%.³⁴

Limitations

Several limitations of the current study warrant discussion. First, this was a retrospective, observational cohort study ascertaining factors associated with initiation of prescribing SGLT2i antihyperglycemic medications. Despite the inclusion of patient- and prescriber-level covariates in the multilevel modelling, 36% of unexplained variation remained in prescription patterns at the provider level, and this variation had a larger impact on prescribing patterns than did patient characteristics. Factors not measured in this study that may influence prescriber behaviour include patient and prescriber beliefs, patient care coordination and contact between primary care and specialists, patient-level socioeconomic status markers, involvement of multidisciplinary teams, and laboratory data, including hemoglobin A1c, albuminuria, and estimated glomerular filtration rate. Second, our study assessed the patterns of initial prescription of antihyperglycemic agents. We acknowledge that some variation may exist in the sequencing of SGLT2i prescriptions. Indeed, 4.2% of patients without an initial SGLT2i prescription were prescribed an SGLT2i at a later date. However, we chose to focus on initial SGLT2i prescriptions to minimize survival bias in our analyses. Third, as this study was quantitative, we were not able to assess why prescribers chose to prescribe or not prescribe SGLT2is. Finally, our study period does not include the years in which indications for SGLT2i were expanded to encompass heart failure and chronic kidney disease in nondiabetic populations. Although an expectation of a substantial increase in uptake of SGLT2i following these guideline changes may be reasonable, prior evidence shows a persistent gap between recommendations and real-world practice. As a corollary from a different clinical context, angiotensin receptor–neprilysin inhibitors (ARNIs) are a class I recommendation in patients with heart failure with reduced ejection fraction³⁵; yet, in a Danish population-based cohort of 42,124 eligible patients, ARNI uptake was only 8.8% over 5 years after publication of guideline recommendations.³⁶ Although the expanded indications for SGLT2 inhibitors are more recent, other population-based studies similarly report low uptake rates.^23,37, 38, 39 Taken together, this evidence suggests that our findings remain relevant, despite exclusion of the most recent indications, as the barriers to SGLT2 inhibitor prescribing are likely to persist.

Conclusion

Significant variation is present in initiation of SGLT2i medications, driven by prescriber practice. Family physicians are responsible for the majority of SGLT2i prescriptions, and they represent a key provider group that could help reduce the gap between prescribing practices and guideline recommendations, particularly among older patients with cardiovascular risk factors who may derive the most benefit from these medications. Continuing medical education and/or electronic medical record practice alerts may help alleviate therapeutic inertia and increase uptake of cardiorenal protective medications.

Disclaimer

This study is based in part on data provided by Alberta Health. The interpretation and conclusions contained herein are those of the researchers and do not necessarily represent the views of the Government of Alberta. Neither the Government of Alberta nor Alberta Health express any opinion in relation to this study.