Inpatient Hypoglycemia: The Challenge Remains

Abstract

Hypoglycemia in inpatients with diabetes remains the most common complication of diabetes therapies. Hypoglycemia is independently associated with increased morbidity and mortality, increased length of stay, increased readmission rate, and increased cost. This review describes the importance of reporting and addressing inpatient hypoglycemia; it further summarizes eight strategies that aid clinicians in the prevention of inpatient hypoglycemia: auditing the electronic medical record, formulary restrictions and dose-limiting strategies, hyperkalemia order sets, electronic glucose management systems, prediction tools, diabetes self-management, remote surveillance, and noninsulin medications.

Hypoglycemia in inpatients with diabetes is the most common complication of diabetes therapy. Insulin remains the preferred treatment of hyperglycemia (blood glucose [BG] ≥ 140 mg/dL) in both critically and noncritically ill patients with and without diabetes.¹

The purpose of this review is to give an update on the burden of inpatient hypoglycemia and summarize newer strategies aimed at reducing inpatient hypoglycemia. PubMed and Google Scholar were searched (“hypoglycaemia”[All Fields] OR “hypoglycemia”[MeSH Terms] OR “hypoglycemia”[All Fields]) AND (“inpatients”[MeSH Terms] OR “inpatients”[All Fields] OR “inpatient”[All Fields]) from January 2015 to March 2020. Relevant articles published in English were identified and included in this review.

Definitions

The International Hypoglycemia Study Group defined three levels of hypoglycemia for the purpose of standardized reporting. Level 1 hypoglycemia or threshold hypoglycemia is defined as BG level below 70 mg/dL but ≥54 mg/dL. At this level increased monitoring and ingestion of carbohydrate should be considered. Level 2 hypoglycemia or clinically significant hypoglycemia is defined as a BG <54 mg/dL. Below this level, neurogenic and neuroglycopenic symptoms normally begin to occur. However, with poorly controlled diabetes or recurrent hypoglycemia, the symptomatic threshold level can shift, and symptoms can occur at higher and lower values. Level 3 hypoglycemia or severe hypoglycemia (SH) is defined as BG <40 mg/dL in hospitalized patients.²

The Burden of Hypoglycemia

Inpatient hypoglycemia is common; however, the reported prevalence and incidence varies depending upon the definition used in each study and the level of glycemic control. Studies using tight control (BG 80-100 mg/dL) in critically ill patients have shown an incidence as high as 45%.³ A recent study reported the prevalence of threshold hypoglycemia and SH to be 10.1% and 1.9%, respectively.⁴

Inpatient hypoglycemia has been associated with negative health outcomes including increased mortality. This association is higher when hypoglycemia is severe.⁵ A 2.6% short-term mortality was found among hypoglycemic admissions.⁶ SH increased the risk of acute coronary syndrome, especially in adults aged 70 or greater and in the first 10 days after the SH event in a large cohort study.⁷ Frequent exposure to hypoglycemia might increase the risk of stroke.⁸ Hypoglycemia was also associated with in-hospital falls.⁹

A large retrospective study showed that 1.1% of patients with type 2 diabetes (T2D) using basal insulin and 3.2% using basal bolus insulin are hospitalized for SH. In contrast, 4% of patients with type 1 diabetes (T1D) were hospitalized for SH. Up to 50% of patients with T2D and 30% of patients with T1D admitted for SH were hospitalized again underscoring the higher costs and higher readmission risk associated with these hospitalizations.^10,11 A calculator predicting inpatient death and 30-day readmission after an admission for hypoglycemia is in development.¹² Hypoglycemia also increases length of stay, even after accounting for the number of glucose tests performed; a recent meta-analysis showed inpatients exposed to hypoglycemia had a 4.1-day longer hospital stay.^13,14 With the goal of providing safer care for inpatients with diabetes, the Centers for Medicare and Medicaid have designated SH as a never event and a safety measure of hospital harm for 2019.

Identifying Inpatient Hypoglycemia

Monitoring BG is an essential component of diabetes management and the preferred strategy to prevent inpatient hypoglycemia. While the majority of patients report symptoms with hypoglycemia, up to 44% of episodes can be asymptomatic, especially in the elderly and those with lower admission A1c levels.¹⁵ The frequency and accuracy of BG monitoring is extremely important; harm can occur when insulin is titrated based on an incorrect value.¹⁶ The recommended monitoring frequency is detailed in Table 1. Portable blood glucose meters (BGMS) are often used because they can provide more frequent measurements than laboratory samples and allow for real-time treatment decisions. In the United States, the Food and Drug Administration (FDA) has specific criteria for hospital meters; 95% of all point of care (POC) measurements should be within ±12 mg/dL of the comparator method for glucose concentration below 100 mg/dL and ±12.5% above 100 mg/dL. They should also accurately measure BG levels down to 10 mg/dL and up to 500 mg/dL. Interference and erroneous measurements can result from physiological and pathological factors like hypotension, hypoxia, or dehydration; medications like acetaminophen; and substances like maltose, dextrose-containing fluids, and even food residue in the hand. It is key to be familiar with the limitations of the BGMS utilized. Verification of hypo- and hyperglycemia values to exclude error can be done based on clinical judgment either by repeating a POC test or by ordering laboratory testing. Hypoglycemia and extremely hyperglycemic values are more likely to be deleted and repeated by nurses. A retrospective review showed that only 0.8% of POC BG tests are repeated; however, using this strategy reduced reporting error and improved documentation of SH by as much as 40%.¹⁷

Table 1.

Frequency of BG Monitoring.¹

NPO	Every 4 h
Eating meals	QID before each meal and at bedtime
Tube feeding	Every 4 h
Total parenteral nutrition	Every 4 h
Intravenous insulin infusion	Every 1 h

BG, blood glucose; NPO, nil per os; QID, four times a day.

Prevention of Hypoglycemia

Minimizing hypoglycemia through risk factor reduction should be the priority as summarized in Table 2.^18-21 A target glucose of 140-180 mg/dL is currently recommended for both critically ill and noncritically ill patients. More stringent targets may be appropriate if they can be achieved without hypoglycemia.¹ Antecedent hypoglycemia has been shown to blunt symptom and counter-regulatory hormonal responses to subsequent hypoglycemia and remains an excellent predictor of subsequent episodes of hypoglycemia.¹⁸

Table 2.

Risk Factors for Inpatient Hypoglycemia.^18-21

Patient factors

Advanced age Decreased renal function Low HbA1c High HbA1c in critically ill patients Long duration of diabetes Impaired hypoglycemia awareness Poor appetite/poor meal intake Chronic liver disease

Dosing and administration factors

Failure to adjust home diabetes regimen Aggressive diabetes management High basal dosing/basal only regimens Mismatch of POC BG testing with insulin dosing and meal delivery Inadequately addressed antecedent hypoglycemia Failure to account for dextrose-containing fluids Frequent insulin dosing and stacking of insulin action Failure to address changes in PO status Failure to address changes in steroid doses Use of nonstandard therapies: SU, premixed insulins, and concentrated insulins. Use of correctional insulin overnight

BG, blood glucose; HbA1c, hemoglobin A1c; PO, per os; POC, point of care; SU, sulfonylurea.

Standardized reporting, order sets, and hypoglycemia protocols are highly effective to reduce inpatient hypoglycemia.^18,22,23 The widespread use of electronic medical records (EMR), continuous subcutaneous insulin infusion (CSII) pumps, and continuous glucose monitoring (CGM) technology has led to renewed efforts to prevent hypoglycemia by developing automated tools, predictive scores, and remote surveillance. Eight strategies are reviewed below.

Auditing the EMR

To improve work processes and standardize care, the EMR can be a powerful tool to elucidate root causes of hypoglycemia. An institution chose to incorporate an automated survey in the EMR for nurses to record root causes of hypoglycemia (BG <70 mg/dL) in real time. They found that insulin dosing and nutrition were the two most common factors associated with hypoglycemia. After a brief targeted education, 642 patients postintervention were compared with 566 historical controls showing that threshold hypoglycemia decreased from 2.3% to 1.5% and recurrent hypoglycemia decreased from 5.7% to 2.2%.²⁴ Auditing the EMR can also help identify processes in need of improvement. A 24-hour hypoglycemia report with all episodes of BG <50 mg/dL was generated by the EMR and reviewed by a diabetes nurse specialist. A detailed root cause analysis led to change in three main work processes over four years: delaying evening snack delivery; standardizing meal delivery timing; and reducing the time lag between POC BG, insulin administration, and meal delivery. This process reduced the rate of BG <50 mg/dL by 50% to a rate ranging from 3.21% to 3.56% per 1000 patient days for the year after the last process improvement was implemented.²⁵ In critically ill patients, real-time evaluation of all episodes of hypoglycemia (BG <60 mg/dL) by nurses significantly decreased the percentage of patients with hypoglycemia among those without diabetes (from 6.15% to 3.78%) and with diabetes (from 13.14% to 7.23%); glucose variability also decreased in both groups.²⁶

Formulary Restrictions, Dosing Calculators, and Dose-Limiting Strategies

Restrictions for the use of premixed insulin and concentrated insulin preparations including regular U-500 insulin, U-200 insulin, and U-300 insulin, and sulfonylureas have been shown to decrease the risk of hypoglycemia.^18,27 Restricting inpatient ordering of high glargine doses of ≥0.5 units/kg to the endocrine team decreased the rate of inpatient hypoglycemia per admission and per BG measurements. Preintervention hypoglycemia rate was 3.4% per BG measurements and decreased to 2.3%; high dose glargine orders decreased from 5% to 0.3%.²⁸ Hypoglycemia also decreased by 50% when a recommended weight-based basal dose was added to the insulin order set at a single institution.²⁹ A recent publication describes the development of a clinical decision support tool incorporated into the EMR, to help physicians start a basal bolus regimen based on weight, home regimen, prior 24-hour insulin requirements, or prior insulin infusion requirements.³⁰

Hyperkalemia Order Sets

Iatrogenic hypoglycemia after administration of insulin for hyperkalemia (serum potassium >5.1 mEq/dL) may occur in 8.7%-17.5% of all patients. Specially at risk are those with impaired renal function, baseline glucose level, and high dose insulin.³¹ To address this problem, a study looked at weight-based dosing of 0.1 units/kg with a maximum dose of 10 units vs the standard 10 units. Potassium lowering was similar in both groups; however, the weight-based group had a reduction of hypoglycemia by 50%.³² A second study that utilized a hyperkalemia order set with additional clinical variables such as renal function and baseline glucose level have showed to further decrease hypoglycemia by 10%. The majority of the hypoglycemic events occurred within three hours post-treatment.³¹

Electronic Glucose Management Systems

Dosing calculators or electronic glucose management systems (eGMS) are decision support systems that provide automated workflow support. Available eGMS include GlucoStabilizer, Glucotab, Glucommander, and EndoTool. They have shown to decrease hypoglycemia in both critically ill and noncritically ill patients.^33,34 They can be useful for the management of diabetic ketoacidosis (DKA); a recent study showed eGMS resulted in no patients experiencing hypoglycemia vs 10 patients who did in the group treated with a protocol-directed provider-guided insulin dose adjustment. The time to resolution of DKA was similar between groups. The eGMS group required significantly less insulin (59.2 units vs 101.2 units).³⁵ Neubauer et al reported that recommendations provided by the eGMS were generally accepted by nurses and physicians.³⁶ A small pilot study with eGMS and CGM technologies showed decreased time spent in hypoglycemia in noncritically ill patients with T2D. The study only included 30 patients but was innovative for its use of CGM and glargine U-300 insulin in inpatients.³⁷ At one large hospital, the application of an eGMS increased the use of basal bolus insulin from 5% to 95% with reduction of hypoglycemia by 21% and SH by 50% highlighting the role of the system to overcome clinical inertia.³⁸ A recent review found that overall eGMS limit the occurrence of hypoglycemia while increasing the time spent at target BG levels.³⁹ Despite the positive evidence, larger and more disease-specific studies are needed to examine cost effectiveness.

Prediction Tools

Leveraging technology to predict inpatient hypoglycemia is an attractive approach that can further minimize inpatient hypoglycemia. Several predictive tools have been developed that translate patient characteristics into a score.⁴⁰ The score is then used to identify patients at risk of hypoglycemia with the goal of implementing a real-time intervention that prevents the hypoglycemic event. The first prediction tool by Tobin et al used an informatics alert in real time that identified patients at risk of SH based on several clinical characteristics including low weight, decreased creatinine clearance, basal insulin doses greater than 0.25 units/kg, or basal only insulin regimens. The alert was triggered when a BG level was below 90. The alert when implemented by trained nurses on medical floors showed a 68% decrease in episodes of SH with a sensitivity of 75%; the sensitivity to predict hypoglycemia at a higher threshold (≤60 mg/dL) was 54%.⁴¹ Using another cohort of more than 128 000 insulin-treated patient days, representing 18 000 patients, Mathioudakis et al developed and validated another predictive tool using additional variables like mean BG, nadir BG, steroid use, sex, T1D, and T2D. The model performed with a sensitivity of 74.6% and specificity of 78.5 for threshold hypoglycemia, and sensitivity of 81.9% and specificity of 78.6% for clinically significant hypoglycemia. The positive likelihood was 3.5, which corresponds to around 23% increased probability compared to the prior published model. Data on effectiveness of this model are not available.⁴² More recently, a scoring tool developed from a smaller cohort using only five clinical variables was able to predict threshold hypoglycemia with a sensitivity of 85% and specificity of 32%.⁴³

Lastly, pattern recognition software is another emerging strategy. This would allow a software to predict hypo- and hyperglycemic patterns with greater accuracy than by simple review of BG data to help clinicians adjust insulin therapy.⁴⁴

The Increasing Role of Diabetes Self-Management

It is estimated that around 60% of individuals with T1D use an insulin pump (CSII) to manage their diabetes. Its use is also increasing among patients with T2D. Insulin pumps allow patients to adjust their insulin based on their meal intake, activity, and BG level. One of the major advantages of this technology when compared with basal bolus insulin is reduction in hypoglycemia. The adoption of CGMs with or without CSII has also improved glycemic control and reduced hypoglycemia in the outpatient setting. People with diabetes are hospitalized at a higher rate than people without diabetes, which translates to more patients presenting to hospitals wearing CSII and CGMs. Data in the inpatient setting are limited to observational and retrospective studies, but appears to suggest that with appropriate patient selection, the use of CSII can be safe and might improve severe hyper- and hypoglycemia. It also improves patient satisfaction. Whenever CSII is discontinued, a basal bolus regimen needs to be instituted to avoid diabetic ketoacidosis. Hospitals need to have clear policies and procedures to guide patients and staff in place. It is strongly encouraged to involve endocrinologists early in the care of these patients.⁴⁵

The use of CGMs in the hospital is an attractive area of research. A consensus statement published in 2017 concluded that there is insufficient evidence to use CGMs to dose insulin; however, they could have an additional role in preventing inpatient hypoglycemia by detecting the fall of glucose before it reaches the level of hypoglycemia and alert the hospital staff. A policy to guide documentation, calibration, and staff response should be in place.⁴⁶ The integration of CGM data to the hospital workflow was tested using remote downloads at nursing stations at one institution. The study included 13 insulin-treated patients, and the rate of hypoglycemia was overall low with six episodes in the group receiving standard care vs two episodes in the intervention group. Despite the difference not being significant, it showed that remote transmission of CGM data can be successful and a larger trial is underway.⁴⁷

Inpatient closed-loop systems represent another highly attractive option to deliver insulin in the inpatient setting. They integrate CSII, a CGM, and a control algorithm that directs insulin delivery on the basis of real-time sensor glucose measurements. Advantages include the decreased need for POC BG monitoring and decreased workload for hospital staff. At this time, evidence of their potential safety and utility is emerging. An open label randomized clinical trial with 136 adults with T2D were assigned to either closed loop or conventional CSII until discharge for up to 15 days. There was better glycemic control in the closed-loop group; no difference in hypoglycemia was found. Neither group had any episodes of clinically significant hypoglycemia or SH. The study was unique as it also included patients on hemodialysis.⁴⁸ A smaller trial with a post-hoc analysis showed increased time in target in the patients on hemodialysis.⁴⁹

Remote Surveillance

The implementation of dynamic systems that allow for remote inpatient glycemic management is emerging as an effective strategy to reduce hypoglycemia. Rushakoff et al implemented a daily automated report in the EMR for patients with two or more BGs ≥225 mg/dL or ≤70 mg/dL in the previous 24-hour period. A note with recommendations by a diabetes specialist was entered in the patient’s chart. This intervention was highly accepted and decreased both hyper- and hypoglycemia. Hypoglycemia decreased by 36%.⁵⁰ A similar intervention by Sheen et al called “electronic dashboard” consisted of a daily automatic report in the EMR accompanied by virtual glycemic management recommendations from endocrinologists. At three years, hyperglycemia decreased by 25% and hypoglycemia by 45% in this observational single-institution study.⁵¹

Noninsulin Medications

Noninsulin therapies have traditionally been discontinued in the inpatient setting due to their slow onset of action, inability to predict their effect, and potential side effects.¹ Incretin-based agents such as glucagon-like peptide-1 (GLP-1) agonist and dipeptidyl peptidase-4 inhibitors carry a low risk of hypoglycemia and may be a safe alternative to complex insulin regimens in the inpatient setting. The safety and efficacy of sitagliptin was explored in a multicenter open label randomized clinical trial with 279 patients. Patients received basal bolus vs basal plus sitagliptin during their hospital stay. There were no statistically significant differences in glucose control or hypoglycemia. A trend to less hypoglycemia was noted in the basal plus sitagliptin group; however, the trial was not powered for this outcome. Sitagliptin showed to be a safe and less labor-intensive alternative.⁵² The use of linagliptin was studied in an open label multicenter randomized trial of patients undergoing noncardiac surgery. Patients were randomized to linagliptin vs basal bolus therapy, and both groups received correctional insulin as needed. The linagliptin group had an 86% relative risk reduction in hypoglycemia; mean daily glucose was slightly inferior in the linagliptin group (171 mg/dL ± 41.4 vs 158.4 mg/dL ±41.4).⁵³

There are few trials studying the inpatient use of GLP-1 agonists. An open label study from India with 123 individuals in the intensive care unit did show a lower rate of hypoglycemia when liraglutide 1.2 mg was used.⁵⁴ A pilot study with 150 patients looked at exenatide twice daily alone, exenatide plus basal and basal bolus. Eligible patients were those requiring less than 0.5 units/kg of insulin per day. Exenatide plus basal was as effective as basal bolus. Hypoglycemia was similar among the groups. Nausea was more common in the exenatide groups (10% vs 2%).⁵⁵ An ongoing trial is exploring the use of liraglutide 0.6 mg the day prior to cardiac surgery and 1.2 mg just prior to surgery vs placebo. The primary endpoint is intraoperative insulin requirements.⁵⁶

The use of the sodium-glucose cotransporter 2 inhibitors (SGLT2i) in the inpatient setting is convenient as once daily oral medications and may lower hypoglycemia risk. Four agents are approved by the FDA: canagliflozin, dapagliflozin, empagliflozin, and ertugliflozin. A number of cardiovascular outcome trials have demonstrated renal and cardioprotective properties in addition to lowering glucose.⁵⁷ They are recommended as first-line therapy in people with diabetes and cardiovascular disease. However, there are no published studies on the use of SGLT2i in the acute phase of cardiovascular events.⁵⁸ An ongoing trial with empagliflozin is looking at heart rate variability after acute myocardial infarction.⁵⁹ Moreover, significant side effects are of concern, specifically the risk of euglycemic diabetic ketoacidosis which in the majority of cases is severe and occurs in the inpatient setting.⁶⁰ It is thought that SGLT2i favor ketogenesis by promoting gluconeogenesis and lipolysis. Inpatients might be at increased risk for this complication as they are more likely to meet certain conditions: severe infections, dehydration, ketogenic fasting, and undergo surgical procedures. At present, these agents should be discontinued three days prior to an elective procedure and immediately in emergent procedures. Caution of its use after bariatric surgery is advised because of the rapid weight loss and carbohydrate restriction in this population. Further evaluation in appropriately designed inpatient clinical studies is needed before implementing them in the inpatient setting.⁵⁸

Conclusion

Preventing inpatient hypoglycemia is a monumental task that requires a series of structured interventions and process improvements with the goal of standardizing hospital care delivery. The interventions described in this review underscore the development, acceptability, and adoption of diabetes technology in the inpatient setting. It also highlights a paradigm shift from traditional reactive approaches often triggered by a hypoglycemic event to more sophisticated proactive interventions and the increased role of patient self-management in the inpatient setting. Finally, more data on the safety and efficacy of noninsulin medications are emerging. Future research should focus on showing improved health outcomes.