Compared to commercially insured patients, Medicare advantage patients adopt newer diabetes drugs more slowly and adhere to them less

ABSTRACT Aims To compare rates of use and adherence for newer versus older second‐line diabetes drug classes in commercially insured, Medicare Advantage and dual‐eligible (covered by both Medicare and Medicaid) patients. Materials and Methods Longitudinal cohort study using insurance claims data from 1/1/2012 to 12/31/2016 to identify patients with a first prescription, after metformin, of a second‐line diabetes drug (eg sulphonylurea, DPP‐4 inhibitor, thiazolidinedione, SGLT‐2 inhibitor or GLP‐1 receptor agonist) and to estimate their adherence to that drug class. Univariate analysis and multivariable logistic regression were used to examine the association between insurance type and use of each drug class, and between insurance type and adherence to each drug class. Results The study population included 96,663 patients. Trends in drug use differed by insurance type. For example, sulphonylurea use declined among the commercially insured (from 46% to 39%, p < .001) but not among Medicare Advantage or dual‐eligible patients. Patterns of adherence also differed between insurance groups. For example, compared to commercial insurance, Medicare Advantage was associated with higher adherence to sulphonylurea (odds ratio [OR] 1.32, 95% CI 1.21–1.43)) but lower adherence to SGLT‐2 inhibitors (OR 0.43 (95% CI 0.33–0.56)). Conclusions This study finds differences in utilization and adherence for diabetes drugs across insurance types. Older medications such as sulphonylureas appear to be more used and better adhered to among Medicare Advantage recipients, while the opposite is true for newer medication classes. These findings suggest a need to personalize selection of diabetes drugs according to insurance status, particularly when adherence needs optimization.


| INTRODUC TI ON
New drugs for treatment of type 2 diabetes mellitus (T2DM) have proliferated over the past 15 years, with the introduction of glucagon-like peptide 1 (GLP-1) receptor agonists, dipeptidyl peptidase-4 (DPP-4) inhibitors and sodium-glucose transport protein 2 (SGLT-2 inhibitors). [1][2][3][4] These agents were originally promoted for having fewer risks of hypoglycaemia and weight gain than older diabetes drugs. The newer diabetes drug classes have also benefited from large recent clinical trials showing that SGLT-2 inhibitors and GLP-1 receptor agonists prevent major adverse cardiovascular events and progression of kidney disease. 1,2 Professional guidelines increasingly encourage use of the newer medications as preferred second-line therapy (after first-line use of metformin) over older agents (ie sulphonylureas, insulin and thiazolidinediones). 1,2 While adoption of newer diabetes drugs has been significant, sulphonylureas remain the most commonly used second-line agent overall. 5 Concerns about contraindications, safety, tolerability and cost all have the potential to hasten or slow the adoption of newer diabetes drugs in specific populations. 3 For example, since the newer diabetes drugs have no generic equivalents and remain relatively expensive, professional guidelines note that older drug classes such as sulphonylureas and thiazolidinediones may be preferred in those patients with limited financial means, since substantial empirical evidence supports that high cost may translate into poor adherence. 1,6-10 At the same time, the increased risk of hypoglycaemia in older adults has led some providers to recommend adopting newer drugs more quickly in that population. 11,12 These competing considerations make it difficult to predict whether, and how quickly, newer diabetes drugs will displace older drugs (especially sulphonylureas) in any given population. For example, commercially insured patients might be expected to have generally higher incomes and lower cost sharing than Medicare patients, which could drive more rapid adoption of newer medications. 13 On the other hand, the high cardiovascular comorbidity burden and hypoglycaemia risk seen in older patients might drive more rapid adoption of newer drugs in Medicare patients.
This study aimed to compare rates of adoption of and adherence to newer diabetes drugs in commercially insured, Medicare Advantage and dual-eligible (covered by both Medicare and Medicaid) patients, test the hypothesis that adoption of and adherence to newer drugs is lower in Medicare patients compared to commercially insured patients and assess whether differences in utilization are explained by differences in out of pocket (OOP) drug cost.

| Population
This longitudinal study used the Health Care Cost Institute (HCCI) database, which merges de-identified claims data from commercial insurance carriers in the United States. 14  these drugs were excluded from most analyses due to the challenges in calculating adherence to insulin and small cohort sizes for meglitinides and alpha-glucosidase inhibitors.
The study cohort was restricted to adult patients with: at least one metformin prescription <90 days prior to the index date; ≥1 year of baseline data prior to the index date with no use of diabetes drugs other than metformin; a non-missing insurance status; a baseline diagnosis of diabetes mellitus based on the presence of at least one ICD-9 or ICD-10 code for diabetes and an identifiable 5-digit zip code of residence. Patients were only included if initiating a single non-metformin (ie second-line) diabetes drug; therefore, patients beginning multiple new diabetes drugs on the same day were excluded. Patients were also excluded if the initial prescription for the second-line drug had >30-day supply or if a mail-order pharmacy was used, on the assumption that such prescriptions were less likely to be true incident use.
For analyses taking adherence as the outcome, patients were also required to have ≥1 year of post-index follow-up data.

| Economic and demographic covariates
Covariates were assessed for each patient using data from the year prior to the index date and included: age, sex, insurance type (com-

| Prescription-related covariates
The study used two cost measures that were associated with each prescription claim: the calculated sum of payments by both patient and insurer (total cost); and the sum of out of pocket (OOP) costs.
Costs reported and analysed are for the initial (index) prescription of the second-line drugs.

| Outcomes
Adherence was defined as the proportion of days covered (PDC) with drug supply during the first year of follow-up, and 'adequate secondary adherence' (defined as fewer than 20% of days without drug supply during the first year of follow-up). 16 These metrics were calculated using the AdhereR package in R. 17

| Analysis
Baseline population characteristics and unadjusted estimates of the outcome rates across all variables were summarized using means, medians, and proportions, and chi-square or t tests as appropriate to calculate statistical significance. Rates of use of different drug classes over time were described. Logistic regression with multivariable adjustment was used to examine the effect of insurance type and other variables on the odds of a patient receiving sulphonylurea as opposed to a newer antidiabetes medication (GLP-1 receptor agonist, SGLT-2 inhibitor or DPP-4 inhibitor) after restricting the cohort to patients receiving those exposures. Logistic regression with multivariable adjustment was also used to generate odds ratios for the effect of insurance type and other variables on adherence to each medication class. In the primary analysis, OOP cost was excluded from this model as a potential mediator. In secondary analysis, it was included to assess for potential mediation.
In sensitivity analyses, 1:1 propensity score matching was used in place of logistic regression for all multivariable analyses. In additional sensitivity analysis, inclusion criteria were relaxed to permit initial prescriptions >30 days and the use of mail-order pharmacy in the initial prescription, initial prescriptions not preceded by a diabetes diagnosis or individuals without zip code level data. Finally, for adherence analyses, the required duration of follow-up was shortened from 12 months to 6 months.

| RE SULTS
After application of inclusion/exclusion criteria, the study population  Table 1).
In analyses of adherence, requiring one year of follow-up data and excluding insulin users, cohort size was further reduced to 76,359 individuals. The cohort was evenly distributed by sex except that the majority (64%) of dual-eligible beneficiaries were female. Age distribution differed by insurance type, with Medicare Advantage patients being older. Medicare Advantage and dual-eligible patients also had higher rates of comorbidities, and greater prescription drug use at baseline. Second-line agent use varied by insurance type, with GLP-1 receptor agonists and SGLT-2 inhibitors used most frequently by commercially insured patients and sulphonylureas used most frequently by Medicare Advantage patients (Table 1).
Over time, rates of sulphonylurea use declined among the commercially insured (from 46% to 39%, p < .001) but not among  Table 2). Older age was independently associated with greater sulphonylurea use, while residence in a higher-income zip code was independently associated with less sulphonylurea use.
Adherence patterns varied by insurance type, with adherence to newer agents higher among commercially insured and dual-eligible patients and lower among Medicare Advantage patients ( Figure 2 Table 3).
Sensitivity analyses using propensity score matching in place of regression, or relaxing inclusion criteria to include patients without baseline diabetes diagnosis or with an initial prescription for longer TA B L E 1 Baseline characteristics. SMD = standardized mean difference relative to commercially insured group  Table 4).

| DISCUSS ION
This study finds substantial differences in patterns of utilization and adherence for diabetes drugs across patients with different types of insurance. These populations are very different from one another, and the data are insufficient to prove or disprove the hypothesis that drug cost mediates these differences. Nonetheless, this finding has practical implications for research and practice, and opens potential new avenues for research into the comparative safety and effectiveness of newer diabetes drug classes.
From a prescriber perspective, these results suggest that Medicare Advantage patients on average adhere best to older drug classes. The causes of these differences in adherence and drug utilization across insurance types need further study. A secondary hypothesis of this study was that higher levels of cost sharing for expensive newer drugs might mediate lower use by Medicare Advantage patients compared to other types of insurance. These analyses did not support that hypothesis, both because OOP cost of drugs were similar between Medicare Advantage and commercially insured patients, and because inclusion of OOP cost in multivariable analysis as a potential mediator did not eliminate observed differences in adherence.

F I G U R E 1 Trends in initial choice of second-line diabetes by year by insurance type. Missing points denote 0 or very low rates of usage such that showing proportions would create cells with fewer than 11 individuals
However, cost may still play a key role in the differences ob- findings are both highly relevant and in need of replication in data more recent than the end of 2016.
In summary, these descriptive analyses provide actionable information. Adoption of SGLT-2 inhibitors has been far more rapid among commercially insured patients than among Medicare Advantage or dual-eligible patients, and sulphonylureas remain particularly widely used among older and Medicare Advantage patients. These findings have direct relevance to research and clinical practice.

ACK N OWLED G EM ENTS
All authors contributed substantively to this manuscript. JHF is the guarantor and conducted the data analysis. JL, GKD and CEL all contributed to study design and manuscript preparation.