Practice guidelines for managing diabetes in older adults increasingly emphasize an individualized approach for determining appropriate glycemic targets and antidiabetic medication (ADM) regimens [3–6, 23]. When making treatment decisions, clinical guidelines universally recommend assessing older patients’ health status and considering potential associated risks, such as hypoglycemia [3–6, 23]. Our study of 84,720 older adults found that the odds of filling SFU or BI was higher for participants in poor health than those in good or intermediate health, even after adjustment for the most recent HbA1c. Filling these same ADM classes was also more common among other subgroups of older adults. For example, SFU fills increased with advancing age; and BI fills increased with greater numbers of diabetes complications. Collectively, these findings demonstrate that SFU and BI remain widely used in older adults, including those who are particularly vulnerable for experiencing adverse events.
This analysis offers a unique perspective into real-world prescribing patterns for second-line ADM in older adults, with findings that are similar to a recent national analysis of electronic medical records [24]. Our use of administrative claims data enabled capture of medication fills, rather than prescription orders that are available from electronic health records but provide limited evidence about whether patients fill the medication. There is little research investigating the gap between existing practice patterns and current consensus guidelines on ADM use in older adults. The poor health group in our analysis is of particular interest and has received little prior study, given that this population is almost universally excluded from clinical trials. Our findings suggest that SFU and BI are used frequently among older adults in poor health, which conflicts with multiple clinical guidelines recommending less frequent use of these medications among vulnerable older adults. One reason for this finding may be related to these medications’ potent glucose-lowering effects [19]. However, short-term adverse events, including hypoglycemia, are frequently described among older adults in poor health, thereby compromising the future benefit of glycemic control in this group. Providers must consider the time horizon for treatment benefit in achieving glycemic targets and minimizing microvascular and macrovascular complications, such that therapy goals are attainable within patients’ life expectancy, and treatment benefit outweighs risk of harm.
Previous studies have demonstrated that older adults with diabetes often receive more aggressive treatment than is warranted, reaching HbA1c values below recommended targets [11, 13, 16, 17, 25]. Further, prior studies show low rates of medication de-intensification in older adults, highlighting a missed opportunity to reduce over-treatment [14, 15]. Several studies evaluate diabetes treatment according to older adults’ health status. However, prior research has not described medication fills using the consensus framework that categorizes patients based on end-stage comorbid conditions, dementia, and/or residence in long term care [11]. Because providers are accustomed to evaluating these comorbidities and characteristics [18, 26], the diabetes treatment framework originally developed by Blaum et al.[18] may prove useful in guiding development of individualized ADM regimens for older adults. Adherence to this framework, and resulting patient outcomes, should be studied empirically in future research that evaluates its implementation. Our findings demonstrate that, despite consensus guidelines from the AGS, ADA, and Endocrine Society recommending use of this framework, older adults in poor health were often treated with SFU and BI, highlighting the persistent gap that this recommendation aims to address.
This paradox is complex, and understanding it more completely requires further study. However, a few explanations and trends are evident. The enduring and frequent use of SFU therapy in older adults may be related to several factors, including its longstanding availability, prescriber familiarity, patient preference, low cost, and significant glucose lowering effects compared to many newer ADM alternatives [19]. Basal insulin is also an established therapy that may be an appropriate patient-centered treatment, especially for those with significantly elevated HbA1c or contraindications to alternative therapy, such as end stage renal disease [19]. Prior research found that older adults with CKD, heart failure, and cardiovascular disease are less likely to fill SGLT-2 medications than those of younger age and those without such comorbid conditions, despite well documented clinical benefits [27]. It is not known whether this is related to concerns about side effects from newer medication classes among patients or providers, their higher cost, or providers’ greater familiarity and experience with older ADM classes. Future research must evaluate the reasons for infrequent use of newer ADM classes among older adults, and study ways to increase their uptake when clinically indicated.
Our study has notable limitations. Observational research is prone to confounding, which may limit causal inference from the findings. However, this issue is of greater potential concern when examining clinical outcomes rather than prescribing patterns examined in the current study. Hypoglycemia or falls were not examined as clinical outcomes here because these are underreported in claims data, and our group recently published an analysis focusing on glycemic outcomes of ADM using the same data source [28, 10]. Claims data do not routinely include laboratory results, although we were able to examine lab data for approximately one-third of participants. While we were unable to fully explore the relationship between HbA1c level and choice of second-line ADM, the primary findings were unchanged after adjusting for the most recent HbA1c result among those with available HbA1c data. We analyzed data on participants’ health plans, but did not have information about formularies or out-of-pocket costs that are known to influence ADM treatment decisions [29]. We used widely accepted definitions for dementia using diagnosis codes outlined by the Centers for Medicare and Medicaid Services [30]. These definitions may not have captured all participants with this exposure in our study, thereby underestimating the number of older adults in poor health. However, prior research on Medicare beneficiaries found that diagnosis codes are more sensitive for identifying older adults with dementia, which comprises the poor health group in the current study, than those with mild cognitive impairment [31]. Detailed data on functional impairment, which is another component of poor health status, is not available in administrative claims data. Prior studies have found that using diagnosis codes to define functional impairment exhibits poor sensitivity and specificity [32]. Clinical evaluation of functional status includes performance-based testing or survey data, which should be included in future research to capture patients with this exposure most completely.