Burden of Hypoglycemia and Hyperglycemia in Insulin‐Treated Veterans Affairs Nursing Home Residents

Authors: Alexandra K. Lee, Ying Shi, Kasia J. Lipska, Sei J. Lee

Abstract

Background

To avoid potential harms from hypoglycemia, guidelines for diabetes management in nursing home residents recommend less intensive glycemic control. However, it is unknown how often hypoglycemia and hyperglycemia co‐occur in the same resident, which may present challenges for deintensification of diabetes treatment.

Methods

We conducted a cross‐sectional study of insulin‐treated Veterans Affairs nursing home residents with diabetes aged ≥ 65 years from 1/1/2016 to 9/30/2019 with a nursing home stay ≥ 7 days. Residents missing fingerstick glucose measurements during the first 7 days were excluded. We classified insulin use as basal insulin only, bolus insulin only, or a combination of basal and bolus insulin. We examined the prevalence of fingerstick‐detected hypoglycemia (< 54 mg/dL, 54–69 mg/dL) and hyperglycemia (250–299, 300–349, 350–399, ≥ 400 mg/dL) overall and stratified by type of insulin.

Results

Among 12,031 insulin‐treated residents, the mean age was 74.4 years, 98% were male, and 22% were non‐White. Most residents (n = 7176, 59.6%) were treated with a combination of basal and bolus insulin, 31.8% (n = 3829) used bolus insulin alone and 8.5% (n = 1026) used basal insulin alone. During the first 7 days of the nursing home stay, 5730 (48%) had hyperglycemia ≥ 250 mg/dL alone, 862 (7%) had hypoglycemia < 70 mg/dL alone, 1488 (12%) had both hyperglycemia and hypoglycemia, and 3951 (33%) had neither hypoglycemia nor hyperglycemia. Residents on a combination of basal and bolus insulin were more likely to have hyperglycemia ≥ 400 mg/dL (10.2% vs. 3.6% for bolus insulin alone and 1.6% for basal insulin alone, p < 0.001) and to have hypoglycemia < 54 mg/dL (8.4% vs. 2.9% for bolus alone vs. 5.9% for basal alone, p < 0.001).

Conclusion

Nearly two‐thirds of nursing home residents with hypoglycemia also had hyperglycemia. Efforts to de‐intensify diabetes treatment in nursing homes will need to address the high burden of hyperglycemia by tailoring the timing and type of insulin to minimize hypoglycemia while also not worsening hyperglycemia.

Introduction

There is abundant evidence that hypoglycemia in older adults is associated with numerous adverse outcomes, including falls and mortality [1, 2, 3, 4, 5, 6, 7]. To decrease the risk of harms from hypoglycemia, type 2 diabetes guidelines for both community‐dwelling older adults and nursing home residents recommend considering less intensive glycemic control and simplification of complex insulin regimens [8, 9, 10]. However, despite these de‐intensification recommendations that have been in place for over a decade, numerous studies of both community‐dwelling and nursing home residents have shown that diabetes overtreatment is common [11, 12, 13] and that de‐intensification of diabetes treatment remains infrequent [14, 15, 16]. However, few studies to date have studied potential barriers to de‐intensification of diabetes treatment in nursing home residents [16].

One area that has not been previously examined as a potential barrier to de‐intensification of insulin therapy is the frequent occurrence of hyperglycemia, particularly in those with hypoglycemia. Because deintensification of insulin may lead to more frequent episodes of hyperglycemia, clinicians may be reluctant to de‐intensify insulin therapy if patients are still having episodes of hyperglycemia. Conversely, frequent hyperglycemia may lead to increases in insulin, resulting in hypoglycemia. A recent nursing home study using CGMs found that 64% of residents using insulin had ≥ 10% of time with hyperglycemia > 250 mg/dL but did not examine how many of these residents also had hypoglycemia [17]. Additionally, it is unknown what proportion of hyperglycemia in nursing home residents is symptomatic vs. asymptomatic, a critical distinction when making treatment decisions that could increase the risk of dangerous hypoglycemia. Over a decade ago, the Diabetes and Aging Study showed that older adults with HbA1c ≥ 9% had a high risk of severe hypoglycemia [18]. This finding suggests that many poorly controlled patients with hyperglycemia are also at high risk for hypoglycemia. Thus, to clarify potential barriers to diabetes de‐intensification, it is crucial to examine the frequency of co‐occurrence of hypoglycemia and hyperglycemia in this population.

Our objective was to quantify the prevalence and co‐occurrence of fingerstick‐detected hypoglycemia and hyperglycemia among insulin‐treated nursing home residents during the first week following nursing home admission. While the first week following nursing home admission may be a time period with more blood glucose fluctuations due to recovery from acute illness, it is also a critical time during which residents' glucose levels often drive subsequent clinical decision‐making about the type and intensity of glucose‐lowering regimen. Additionally, our prior research has shown that more people receive frequent fingerstick glucose measurements during the early part of the nursing home stay, resulting in more reliable prevalence estimates of dysglycemia [19]. We hypothesized that a majority of nursing home residents with hypoglycemia also experience hyperglycemia, prompting continued use of bolus insulin despite the hypoglycemia risk.

Methods

We used the Veterans Affairs (VA) electronic health records to create a cohort of Veterans aged ≥ 65 years with diabetes, a nursing home stay ≥ 7 days from 1 Jan 2016 to 30 Sept 2019, and any insulin administration during the first week following admission (n = 13,988). We excluded residents without any fingerstick glucose measurements in Days 1 through 7 (n = 1957), for a final sample size of 12,031.

We examined clinical and demographic characteristics including age, sex, race/ethnicity, BMI, HbA1c, and median daily fingersticks. For both BMI and HbA1c, we included the most recent measurement available in the electronic health record, up to 1 year prior to admission. We categorized insulin use over Days 1–7 of the nursing home stay as basal insulin only, bolus insulin only, or both basal and bolus insulin.

Using fingerstick glucose values from Days 1 through 7, we calculated the proportion of nursing home residents with hypoglycemia (any single fingerstick value < 70 mg/dL during Days 1–7) and with hyperglycemia (any single fingerstick value ≥ 250 mg/dL, the recommended upper limit for nursing home residents) [8]. To characterize the severity of dysglycemia episodes, we also examined hypoglycemia < 54 mg/dL and hyperglycemia at 50 mg/dL increments (i.e., ≥ 250 mg/dL, ≥ 300 mg/dL, etc.), as some guidelines recommend interventions at ≥ 300 mg/dL [9].

We used logistic regression to examine baseline characteristics that were associated with dysglycemia during the first week of the nursing home stay. We conducted separate logistic regression models for the outcome of hypoglycemia < 70 mg/dL and for the outcome of hyperglycemia ≥ 250 mg/dL. To determine if there were distinct predictors of the co‐occurrence of hypoglycemia and hyperglycemia, we conducted two separate analyses. First, we restricted the analysis to people with hypoglycemia and conducted logistic regression to identify predictors of hyperglycemia among people with hypoglycemia. Second, we restricted the analysis to people with hyperglycemia and conducted logistic regression to identify predictors of hypoglycemia among those with hyperglycemia.

Results

There were 12,031 insulin‐treated residents; mean age was 74.4 years, 98% were male, and 22% were non‐White (Table 1). Mean HbA1c was 7.7%, and the median number of fingersticks per day was 3. Most residents used a combination of basal and bolus insulin (n = 7176, 59.6%). Only 1026 (8.5%) of residents used basal insulin only; 3829 (31.8%) used bolus insulin alone. Almost one‐third of residents additionally used other glucose‐lowering medications. Mean HbA1c was highest in those using a combination of basal and bolus insulin (HbA1c 8.1% vs. 7.3% for basal insulin only and 7.1% for bolus insulin only).

Overall, 3951 (33%) had neither hypoglycemia nor hyperglycemia, 5730 (48%) had hyperglycemia alone, 862 (7%) had hypoglycemia alone and 1488 (12%) had both hyperglycemia and hypoglycemia during the first 7 days of the nursing home stay (Figure 1). Thus, among the residents with hypoglycemia, about two‐thirds also experienced hyperglycemia.

When stratified by insulin therapy, there were substantial differences in the proportions of people with hypoglycemia (< 70 mg/dL), hyperglycemia (≥ 250 mg/dL), both, or neither (Figure S1). Among those on basal and bolus insulin, 7.8% had hypoglycemia alone, 17.2% had hypoglycemia and hyperglycemia, and 54.4% had hyperglycemia alone (20.7% had neither hypoglycemia nor hyperglycemia). Among those on bolus insulin only, 4.7% had hypoglycemia alone, 5.0% had both hypoglycemia and hyperglycemia, and 40.2% had hyperglycemia alone (50.1% had neither hypoglycemia nor hyperglycemia). Among those on basal insulin only, 12.0% had hypoglycemia alone, 5.9% had both hypoglycemia and hyperglycemia, and 28.4% had hyperglycemia alone (53.7% had neither hypoglycemia nor hyperglycemia).

Both hypoglycemia and hyperglycemia were more common and more severe among residents on basal and bolus insulin compared to either basal insulin alone or bolus insulin alone (Figure 2). Of note, hyperglycemia ≥ 400 mg/dL was far more common in those on basal and bolus insulin (10.2%) compared to those on bolus insulin only (3.6%) or basal insulin only (1.6%) (p < 0.001 for group comparison). Hypoglycemia was more common and more severe in those on basal insulin alone compared to bolus insulin among residents on basal insulin only, 5.9% had hypoglycemia < 54 mg/dL compared to 2.9% in residents on bolus insulin only (p < 0.001 for group comparison).

In multivariable‐adjusted models, several demographic and clinical factors were associated with hyperglycemia and hypoglycemia. Older age was associated with increased odds of hyperglycemia (odds ratio (OR) per 5 years of age, 1.07, 95% confidence interval (CI): 1.04–1.11) but not hypoglycemia (Table 2). Compared to basal insulin only, the odds of hyperglycemia were greater with bolus insulin only (OR: 1.84, 95% CI: 1.55–2.18) and with basal‐bolus insulin (OR: 4.63, 95% CI: 3.92–5.45); the odds of hypoglycemia were lower with bolus insulin only compared to basal insulin only (OR: 0.49, 95% CI: 0.40–0.61) and higher with basal‐bolus insulin compared to basal insulin only (OR: 1.47, 95% CI: 1.21–1.78). Use of non‐insulin diabetes medications was associated with lower odds of hyperglycemia (OR: 0.90, 95% CI: 0.82–0.98) and greater odds of hypoglycemia (OR: 1.19, 95% CI: 1.07–1.33). Additionally, lower BMI and higher HbA1c were associated with increased odds of both hypoglycemia and hyperglycemia. Predictors of co‐occurring hypoglycemia and hyperglycemia were not notably different from predictors of either dysglycemic event individually (Table S1).

Discussion

Approximately two‐thirds of VA nursing home residents had hyperglycemia ≥ 250 mg/dL during the first week of their nursing home stay; this was true even among nursing home residents with hypoglycemia < 70 mg/dL. Over 45% of nursing home residents using a combination of basal and bolus insulin had hyperglycemia > 300 mg/dL, suggesting that bolus insulin may often be continued to manage hyperglycemia. Residents on a combination of basal and bolus insulin were also most likely to have hypoglycemia < 54 mg/dL, indicating that the risks of harm are greatest in this group, but de‐intensification of insulin therapy may be challenging due to commonly co‐occurring hyperglycemia.

Our results based on fingerstick measurements of blood glucose were largely similar to a recent study using continuous glucose monitors in nursing home residents [17]. Both studies showed that the majority of insulin users in nursing homes have frequent and significant hyperglycemia > 250 mg/dL, and that hyperglycemia is far more common than hypoglycemia [17]. Overall, the similarity in these findings suggests that fingersticks from routine clinical care among residents on insulin may adequately capture population‐level glycemic patterns when compared to continuous glucose monitors. This similarity in findings is highly reassuring, as it means that the routinely collected fingerstick data can be used for surveillance of population‐level trends in dysglycemia among insulin‐treated nursing home residents. However, routinely collected fingerstick data are unlikely to capture population glycemic patterns of residents without insulin treatment, because fingersticks are conducted less frequently and are more likely to be conducted in response to dysglycemia symptoms [19]. Our study additionally displays higher thresholds of hyperglycemia (e.g., > 300, > 400 mg/dL) across different types of insulin therapy, as well as whether hypoglycemia and hyperglycemia occur in the same person. Further studies are needed to determine if the number of hyperglycemia readings from fingerstick data correlates with the percent of time above range from continuous glucose monitors.

In our study, residents on a combination of basal and bolus insulin had the highest rates of both hypoglycemia and hyperglycemia. Basal bolus insulin regimens require careful attention to meal timing, food intake, and insulin dosing, which may be difficult with workforce shortages in the nursing home setting. One potential option to reduce the burden of hypoglycemia and hyperglycemia is to change medication regimens. While some older adults with absolute insulin deficiency due to longstanding type 2 diabetes may require both basal and bolus insulin, it may be possible to reduce or eliminate bolus insulin among those with remaining pancreatic beta‐cell function. One prior single‐arm study in older adults with type 2 diabetes found that simplifying insulin regimens by prioritizing orals and basal insulin reduced bolus insulin use and time spent in hypoglycemia, but there was a non‐significant trend towards increased time in hyperglycemia [20]. Despite being a small (n = 65), single‐arm study without a control group, this study is the scientific basis for the deintensification algorithm in current clinical guidelines [21, 22]. While there is a small but growing number of randomized controlled trials testing different insulin simplification algorithms, most focus on HbA1c as the primary outcome, and none have been conducted specifically among multimorbid older adults [22]. Additionally, several insulin simplification algorithms specifically focus on adding SGLT2 and/or GLP1 while removing bolus insulin [23, 24, 25, 26, 27]; these newer cardiometabolic diabetes medications need more rigorous evidence of safety and effectiveness in nursing home residents [28]. Rigorous randomized controlled trials of insulin simplification algorithms among medically complex older adults are needed to determine whether the use of non‐insulin glucose‐lowering medications and basal insulin alone can adequately control hyperglycemia in nursing home residents.

In our study, only 25% of residents who used a combination of basal and bolus insulin were using other glucose‐lowering medications, indicating there is potential to increase other glucose‐lowering medications while simplifying insulin regimens (in those with type 2 diabetes). The 25% rate of oral medication use in the VA is substantially lower than another study of long‐term care residents from Pandya et.al. in which 45%–50% of those on insulin were also using oral medications [29]. One other area of concern in our study was the modestly high prevalence of sulfonylureas (13%), which also have a high hypoglycemia risk. In this VA cohort, use of sulfonylureas was highest among those with bolus insulin only (20% sulfonylurea use) compared to basal insulin only (16% sulfonylurea use) or basal and bolus (8% sulfonylurea use). While we did not have statistical power to look specifically at sulfonylureas, our results showed that non‐insulin diabetes medications were associated with a modestly increased risk of hypoglycemia and a decreased risk of hyperglycemia. Additionally, use of SGLT2 and GLP1 was rare (1%), as these medications had strict pre‐authorization criteria in the VA during the study time period [30], which was before the new 2020 guidelines recommended SGLT2s and GLP1s as adjunctive cardio‐renal protective therapy [31].

One additional interesting finding was that those on basal insulin only also had more frequent episodes of hypoglycemia than those on bolus insulin alone. This suggests that bolus insulin may be primarily used to reactively correct hyperglycemia while basal insulin is used proactively to prevent hyperglycemia. In recently admitted nursing home residents who may have irregular oral intake, proactive basal insulin may require closer supervision than is practical in the nursing setting. Minimizing iatrogenic hypoglycemia from basal insulin should be a high priority in the nursing home setting [32].

When deciding treatment approaches for residents, it is important to consider that the short‐term harms from hypoglycemia and hyperglycemia are not equal. Hypoglycemia, particularly severe events requiring assistance, may increase the risk of falls and cardiovascular events; in contrast, single (or even repeated) episodes of hyperglycemia may increase polyuria. We lack good‐quality data on what frequency and degree of hyperglycemia may increase potential harms such as poor wound healing and increased risk of infections.

Our study raises important questions about the value of frequent blood glucose measurements in the nursing home setting. If the goal of diabetes management is primarily to avoid symptomatic hyperglycemia and severe hypoglycemic events, do frequent fingersticks help to achieve this or would other approaches, perhaps with less insulin and less monitoring, also suffice? Research is needed on diabetes management approaches that prioritize reduced patient burden (fewer fingersticks and injections) while maintaining patient safety (avoiding severe hypoglycemia or extreme hyperglycemia that could worsen quality of life), following other deprescribing research [33, 34, 35].

The major strength of this study was its reliance on routinely collected daily recordings of glucose fingerstick values and medication administration in an understudied nursing home population. The primary limitation was that our VA cohort is 98% male and therefore not representative of nursing home residents nationwide; subsequent studies should replicate our study in non‐VA nursing homes. Second, we did not distinguish between type 1 and type 2 diabetes; however, prior nursing home studies have shown only 2% of residents have type 1 diabetes [29]. Third, we looked only at the first 7 days of the nursing home admission, a transition time when blood glucose may fluctuate more than subsequent time periods of a nursing home stay. Finally, we could not assess the duration of hypoglycemia and hyperglycemia given the reliance on single fingerstick values.

In conclusion, among nursing home residents on insulin, over two‐thirds who had hypoglycemia also had co‐occurring hyperglycemia during the first week of their nursing home stay. Bolus insulin use is common in VA nursing homes and episodes of significant hyperglycemia may be a barrier to insulin deintensification. Future studies should examine whether responding to hyperglycemia with orals or basal insulin (rather than bolus insulin) can decrease subsequent hypoglycemia without worsening hyperglycemia or other clinical outcomes. Additionally, research is needed to determine what threshold of hyperglycemia negatively impacts clinical outcomes and quality of life for nursing home residents. Clinicians should carefully consider the timing and dosing of insulin, particularly for residents with fluctuating oral intake, to minimize hypoglycemia and hyperglycemia.

Author Contributions

A.K.L. was responsible for study concept and design, data acquisition, analysis and interpretation of the data, and preparation of the manuscript. Y.S. contributed substantially to study design, data acquisition, statistical analysis, and interpretation of the data. K.J.L. contributed substantially to study design and interpretation of the data. S.J.L. contributed substantially to study concept and design, data acquisition, and analysis and interpretation of the data. All authors reviewed the manuscript critically for important intellectual content and provided final approval of the version to be published.

Funding

Dr. Alexandra Lee is supported by NIH grant K01AG073532. Dr. Sei Lee is supported by NIH grant K24AG066998.

Disclosure

The sponsor played no role in study design, statistical analysis, interpretation of results, preparation or review of the manuscript or the decision to submit the manuscript for publication. A.K.L. 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.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information