A National Physician Survey of Deintensifying Diabetes Medications for Older Adults With Type 2 Diabetes

The benefits of glycemic control to prevent diabetes complications take years to accrue and diminish with limited life expectancy (1). Therefore, guidelines from the American Diabetes Association (ADA) and others recommend selecting higher hemoglobin A_1c (HbA_1c) targets for older adults with worse health status defined by multiple comorbidities, functional or cognitive limitations, and limited life expectancy (2). Accordingly, diabetes guidelines recommend deintensifying therapy (i.e., reducing or stopping a diabetes medication) when it can prevent medication-related harms and be achieved within the individualized glycemic target (2). However, deintensification rarely occurs in practice, and glycemic control does not vary by the patient’s health status (3–5). Therefore, we conducted a national survey of U.S. physicians to understand their decisions to deintensify diabetes therapy and select glycemic targets for older adults.

We surveyed 1,950 physicians in general medicine (n = 525), geriatrics (n = 525), and endocrinology (n = 900) identified using the American Medical Association Physician Masterfile (6). Endocrinologists were oversampled because medical specialists had lower response rates in prior surveys (7–10). Trainees and physicians not routinely providing outpatient care to older adults with type 2 diabetes were excluded. The survey was distributed in three mailing waves with a $10 gift card in the first wave and follow-up e-mails and phone calls to nonrespondents (Supplementary Survey Methods). Responses were received between June 2021 and March 2022.

The survey instrument was developed through a participatory action research process (11) involving patients, caregivers, and multidisciplinary physicians. The final survey (Supplementary Material) included three scenarios representing older adults with type 2 diabetes in different states of health (good, complex, and poor) who were taking an intermediate dose of a hypoglycemia-causing medication with HbA_1c below their individualized target (2). Scenario 1 was a healthy 79-year-old woman taking glimepiride with an HbA_1c of 6.3%. Scenario 2 was a 77-year-old man with multiple chronic conditions taking insulin glargine with an HbA_1c of 7.3%. Scenario 3 was a 78-year-old woman with advanced dementia taking glipizide with an HbA_1c of 7.7%. Each scenario started with a base case in which the patient reported no recent hypoglycemia or medication concerns, followed by nine variations with different patient characteristics and preferences. For each variation, physicians were asked the identical question of what change they would make, if any, to the diabetes medication. The survey also asked physicians to select the HbA_1c target they would recommend for older adults in good, complex, and poor health as described using the language in the ADA Standards of Medical Care in Diabetes—2022 (2).

Characteristics were compared across specialties using χ² or Kruskal-Wallis test. We used multinomial logistic regression models (Supplementary Table 1) to determine the association between predictors of interest and physicians’ actions in clinical scenarios, adjusted for physician specialty and accounting for clustering of responses within physicians. Analyses were conducted using Stata 14 statistical software (StataCorp LP, College Station, TX).

There were 445 eligible respondents (Table 1). The response rate was 37.5%, with 38.3%, 30.0%, and 40.8% response rates from physicians in general medicine, geriatrics, and endocrinology, respectively (Supplementary Fig. 1). The most common reason for ineligibility was not providing outpatient diabetes care (Supplementary Table 2). There were no significant differences between the characteristics of respondents and nonrespondents, except specialty (Supplementary Table 3). Physicians of different specialties had similar demographic, but different practice characteristics (Supplementary Table 4). On average, respondents completed 98% of the survey questions.

In the base case of clinical scenarios (patient had no hypoglycemia or medication concerns), physicians predominantly made no change to the hypoglycemia-causing medication (Fig. 1). Physicians deintensified more often for the healthy patient with an HbA_1c of 6.3% than for patients with more complex health with an HbA_1c of 7.3–7.7% (P < 0.001). Physicians switched medications 17% of the time overall, with no significant difference across scenarios. There were statistically significant differences in deintensification and switching by primary specialty (Supplementary Table 5), although the overall pattern of responses was consistent.

In scenario variations (Fig. 2), a history of recent hypoglycemia had the strongest effect on physicians’ actions. Even mild hypoglycemia caused >99% of physicians to modify therapy, predominantly by deintensifying. In variations where the patient had a major hypoglycemia risk factor, physicians were more likely to deintensify medications compared with the base case and less likely to switch. Notably, a substantial minority of physicians (12–27% depending on the scenario) made no medication change for a significant decline in health (Supplementary Fig. 2). Physicians were modestly responsive to patient preferences for less medication or tighter glycemic control (Fig. 2). There were significant differences in physicians’ responses to scenario variations by specialty such that geriatricians were the most likely to stop medications and endocrinologists the most likely to switch (Supplementary Table 6).

Most physicians selected guideline-concordant HbA_1c targets for patients in good health (Fig. 3). An HbA_1c of 7.0% was the most common target, selected by 60% of physicians (Supplementary Table 7). The majority of physicians (90%) selected a higher target for patients with complex health than they did for patients with good health, and 81% selected a higher target for patients with poor health than they did for complex health. However, for patients with complex and poor health, more physicians selected HbA_1c targets below guidelines than selected guideline-concordant targets. Generalists and endocrinologists selected similar HbA_1c targets, while geriatricians were more likely to select higher targets, especially for patients with poor health (Supplementary Table 7).

Physicians who selected HbA_1c targets below ADA guidelines deintensified hypoglycemia-causing medications infrequently in all clinical scenarios (Supplementary Table 8). One-half of physicians who selected a guideline-concordant HbA_1c target for patients with good health deintensified in clinical scenario 1, whereas 23% of physicians who selected a target below guidelines deintensified in this scenario (P < 0.001). For patients with complex health, 7% of physicians who selected a guideline-concordant HbA_1c target deintensified in clinical scenario 2, whereas 2% of physicians who selected a target below guidelines deintensified in this scenario (P = 0.06). For patients with poor health, 27% of physicians who selected a guideline-concordant HbA_1c target deintensified in clinical scenario 3, whereas 9% of physicians who selected a target below guidelines deintensified in this scenario (P < 0.001). Physicians’ HbA_1c targets were not significantly associated with switching diabetes medications (Supplementary Table 9).

In this national survey, we found that most U.S. physicians would not deintensify or switch sulfonylureas or insulin for older adults with complex or poor health and higher HbA_1c levels (7.3–7.7%). This approach does not align with ADA guidelines that recommend deintensification within individualized HbA_1c targets for patients at increased hypoglycemia risk (2). Furthermore, most physicians selected lower HbA_1c targets than guidelines for older adults with complex and poor health, and selecting a lower target was strongly associated with not deintensifying. Therefore, optimizing deintensification decisions will require increasing physicians’ comfort with higher HbA_1c targets for older adults with complex health and limited life expectancy for whom the harms of hypoglycemia-causing medications may outweigh the benefits (5).

The reasons physicians prefer lower HbA_1c targets than recommended in guidelines are unclear and require further study. Medical training may play a role, given our findings that geriatricians were more likely to deintensify medications than other specialties and that endocrinologists were more likely to switch. It was surprising that so few physicians chose to switch hypoglycemia-causing medications given the availability of newer medication classes with low hypoglycemia risk. Physicians’ reluctance to switch may be related to limited insurance coverage for newer medications, which is a significant barrier to their use in clinical practice (12).

Nearly all respondents appropriately deintensified or switched therapy where the patient reported a history of recent hypoglycemia, suggesting that physicians view hypoglycemia history as the most important factor in considering a patient’s hypoglycemia risk. While all clinical scenarios in this survey provided the patients’ hypoglycemia history, in practice, patients often do not report hypoglycemic events to their provider, and many events may be missed (13). Therefore, routine assessment of hypoglycemia history may be critical for promoting appropriate medication change to prevent hypoglycemia.

We found that most physicians deintensified hypoglycemia-causing medications where patients had additional major hypoglycemia risk factors. However, deintensification was not universal, including where patients experienced a serious decline in health, such as initiating hospice. Physicians’ inertia in these cases likely contributes to patient harm, as exemplified by the large number of patients who experience hypoglycemia while receiving hospice care (14).

Strengths of this study include that the survey was designed through collaboration between clinical and community stakeholders and that we surveyed a national sample of the physicians providing the majority of outpatient diabetes care (15). This study also has several limitations. The clinical scenarios included relatively simple diabetes medication regimens and, thus, did not address simplification of complex regimens (2,16). Response rates were <50%, which may have introduced response bias. Geriatricians had a lower response rate than other specialties, although the interpretation of the main findings is the same across specialties.

In conclusion, from this national survey, we found that physicians’ approach to deintensifying and switching hypoglycemia-causing medications did not align with diabetes guidelines because of inertia against modifying therapy and selecting lower HbA_1c targets for patients with complex and poor health. There is a need to change physicians’ practices to promote individualized treatment decisions that can improve the safety of diabetes care for many older adults.

Funding. This study was funded by the U.S. Deprescribing Research Network, which is funded by the National Institute on Aging (NIA) (R24AG064025). S.J.P. was supported by the Johns Hopkins KL2 Clinical Research Scholars Program (KL2TR003099) and the National Institute of Diabetes and Digestive and Kidney Diseases (K23DK128572). N.L.S. was supported by the NIA (K76AG059984). C.M.B. was also supported by the U.S. Deprescribing Research Network (NIA grant R24AG064025) and by NIA grant K24AG056578. N.N.M. was supported by the NIDDK (R01DK125780).

Duality of Interest. C.M.B. received honoraria for writing a chapter on multiple chronic conditions for UpToDate and a chapter on falls in older adults for DynaMed. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. S.J.P. contributed to the study design, data acquisition, analysis, and discussion and wrote the manuscript. R.J. contributed to the data acquisition and background research and reviewed and edited the manuscript. O.T. contributed to the analysis and discussion and reviewed and edited the manuscript. N.L.S., C.M.B., S.H.G., N.N.M., and N.M.M. contributed to the study design, analysis, and discussion and reviewed and edited the manuscript. All authors approved the final version of the manuscript. S.J.P. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this article were presented in poster form at the Society of General Internal Medicine Annual Meeting, Orlando, FL, 6–9 April 2022, and the 82nd Scientific Sessions of the American Diabetes Association, New Orleans, LA, 3–7 June 2022.