Supporting the Use of a Person’s Own Diabetes Technology in the Inpatient Setting Free

Expert groups recommend that cognizant people with diabetes who can safely operate their diabetes devices should be permitted to do so in a supervised inpatient setting (12–17). The American Association of Clinical Endocrinology (AACE) makes the additional recommendation that these devices ideally would be continued in the presence of a knowledgeable family member or with the support of a specialized inpatient diabetes team (14).

With the U.S. Food and Drug Administration’s (FDA’s) allowance for the use of remote patient monitoring in hospitals during the coronavirus disease 2019 (COVID-19) pandemic (18), the American Diabetes Association (ADA) and AACE both included updated recommendations for inpatient CGM use in their guidelines (12,14). These recommendations include continuation of devices used before hospitalization, as well as initiation of devices in the inpatient setting. The ADA’s Standards of Medical Care in Diabetes—2022 acknowledges that inpatient CGM decreases the use of personal protective equipment by hospital staff but states that clear data indicating that CGM improves glycemic or overall hospitalization outcomes are not yet available (12). Small studies suggest that the continuation of insulin pump therapy in the hospital setting is safe within the context of established programs (19–21). In the largest published case series of 136 patients, continuation of insulin pump therapy was associated with fewer severe hypoglycemic (<40 mg/dL) or hyperglycemic (>350 mg/dL) excursions (21). However, adverse events such as infusion site failures and pump malfunctions are possible (21–24).

There is consensus that, whenever possible, individuals with diabetes should be permitted to continue their use of personal diabetes technology devices during hospitalization when they are willing and able to do so and the proper inpatient support is provided (10,12,14–17). Table 1 provides a checklist of patient- and device-specific considerations for the continuation of personal diabetes devices in the inpatient setting.

Although ADA and AACE guidelines and other consensus reports state that the presence of a family member skilled in pump management may facilitate continued use of technology during hospitalization (12–14), family members are not considered to be a suitable alternative at the authors’ institution because of patient safety concerns. Patients who are critically ill and those with rapidly changing clinical status, any change in mentation, poor nutrition intake, or suicidal ideation are inappropriate candidates for inpatient technology use (10,16,17). At the authors’ institution, the diabetes/endocrinology team works closely with psychiatry personnel to determine whether and when therapy should be restarted for patients with suicidal ideation.

In addition to physical condition, assessment should include evaluation of baseline diabetes self-management and technology knowledge (14). Kannan et al. (19) found that 24% (12 of 50) of patients admitted to the hospital were unable to demonstrate vital pump settings, and a variable proportion could not use the bolus calculator (22%), suspend the pump (22%), set a temporary basal rate (44%), count carbohydrates (12%), use back-up insulin in case of pump malfunction (12%), perform correction boluses (14%), or correct hypoglycemia (8%). Some skills such as the ability to accurately enter carbohydrate and glucose information into the pump’s bolus calculator or the ability to suspend the pump are essential for continuation of therapy because patients will need to perform these tasks independently while hospitalized. However, the use of advanced features (e.g., temporary basal rates) may be performed with the support of the diabetes/endocrinology team.

The diabetes device download is a valuable tool for assessing patient technology proficiency, diabetes management patterns, and glycemic management. Downloading data from devices in the inpatient setting can be time-consuming, requires staff training, and often requires approvals and assistance from information technology personnel. Nonetheless, these downloads can provide information for a robust evaluation, which should include the following components.

Time spent in, below, and above the target glycemic range and mean glucose value should be evaluated in concert with device settings. Patients with predominate hyperglycemia in the outpatient setting are unlikely to achieve inpatient glycemic targets during hospitalization without an increase in insulin delivery. Conversely, patients with tight glycemic management may require a decrease in insulin delivery to prevent hypoglycemia and achieve a slightly higher hospital glucose target. Patients with a history of hypoglycemia are more likely to experience hypoglycemia during hospitalization (11). Hypoglycemia frequency, patterns, and severity, as well as the patient’s ability to sense hypoglycemia, should be evaluated, and the presence of hypoglycemia should warrant preventive adjustments by the diabetes/endocrinology team.

Total daily dose (TDD) of insulin, the basal insulin percentage, and device settings should be evaluated. Basal-heavy regimens are associated with a decline in glycemic management and higher rates of hypoglycemia (25,26). Patients on a basal-heavy regimen who experience a change in nutritional status or who are restricted to nothing-by-mouth (NPO) status are at greater risk for hypoglycemia, and preemptive adjustment of basal rates or use of a temporary basal rate may be warranted. An assessment of bolus frequency and basal/bolus ratios may be helpful in this setting. Additionally, the patient’s prandial insulin settings should be evaluated alongside the TDD and bolus delivery history to ensure that these settings do not place the patient at risk for hypoglycemia.

Bolus frequency, use of bolus calculator/overrides, daily carbohydrate intake, and carbohydrates per bolus should be evaluated. For patients not entering carbohydrates and for those not using the pump’s bolus calculator system, their ability to count carbohydrates and to use the pump’s bolus dose delivery system should be assessed.

All forms of diabetes technology, including CGM devices, traditional insulin pumps, and AID systems, present unique challenges in the inpatient setting. Additionally, technology specifications may vary by specific product brand.

CGM provides frequent measurement of interstitial glucose along with glycemic trend data (13). Although the recent clinical application of CGM in the inpatient setting during the COVID-19 pandemic has received considerable attention, the continuation of personal CGM use in the inpatient setting has received less discussion. Patients generally prefer to remain on their CGM system during hospitalization (16); thus, continuation of use may improve patient satisfaction.

It is important to communicate clearly with patients that, while these systems are approved for outpatient use, they are not at present FDA-approved for glucose monitoring or insulin dosing in the inpatient setting. For this reason, point-of-care (POC) blood glucose monitoring should be performed, with CGM serving as an adjunctive technology (16).

Most CGM systems on the market provide real-time alarms for hypoglycemia and hyperglycemia and predictive hypoglycemia alerts (27–30). Adjusting these alarms, especially the hypoglycemia threshold alarm, should be considered for patients at high risk for hypoglycemia.

CGM devices should be removed before any computerized tomography (CT) scan or MRI, before diathermy, or when the sensor site is in the surgical field (14,31). A recent study by Thomas et al. (32) demonstrated no interference in sensor accuracy after radiographic procedures such as X-ray and CT scan; however, confirmation of these findings is needed.

If a patient uses a CGM system that requires calibration, the hospital glucose meter, rather than the patient’s home meter, should be used to limit inaccurate readings (16).

The continuation of personal insulin pump therapy during hospitalization has proven to be safe and has been shown to increase patient satisfaction (24). With traditional insulin pumps, basal insulin is delivered via pre-programmed rates, allowing varied basal insulin dosing throughout the day. Insulin-to-carbohydrate ratios, an insulin sensitivity factor, and glucose target information is programmed into the pump to facilitate precise bolus dosing. Pump users must calculate and enter carbohydrates into the pump. They must also manually enter hospital POC blood glucose values into the pump for correction bolus dosing.

With traditional insulin pump therapy, insulin delivery can be temporarily adjusted to accommodate inpatient insulin requirements by creating a new basal profile or using the temporary basal feature. This feature may be used to reduce insulin while a patient is on NPO status or may be used to increase insulin delivery to accommodate increased insulin requirements associated with steroid administration or stress hyperglycemia. Before surgical procedures, the temporary basal rate can be set to reduce insulin delivery before and during the procedure. The maximum duration for a temporary basal rate varies by insulin pump brand from 12 hours (Omnipod) (33) to 72 hours (Tandem) (34). Patients should be instructed on how to turn the temporary basal rate off when the advanced feature is used.

AID systems use CGM data and automated algorithms to adjust correction boluses and/or titrate basal insulin doses. Unlike traditional insulin pump therapy, the mechanism of insulin delivery can vary considerably by device brand; therefore, brand-specific factors should be considered when determining whether AID should be continued during hospitalization.

There is a lack of available information addressing the insulin delivery differences among AID systems and how they may affect inpatient use. Medtronic (e.g., MiniMed 670G, 770G, 780G) uses a type of machine learning algorithm (proportional integrative differential), which is based on the previous 2–6 days of use, including active insulin time and glucose levels, and a derivation of insulin sensitivity from the insulin requirements in the previous 2–6 days (35,36). In a similar fashion, Insulet’s Omnipod 5 system uses a machine learning algorithm (personalized model predictive control), which is based on the previous 3 days of use, to adjust basal insulin delivery (35,36). In contrast, Tandem Diabetes Care’s t:slim X2 with Control-IQ system features an algorithm (model predictive controller) that is based on the glucose level, active insulin time, pre-programmed basal rates, and insulin sensitivity (35,36). Pre-programmed basal rates and insulin sensitivity can be adjusted in the Control-IQ algorithm, whereas the Medtronic and Omnipod systems do not allow for adjustment of these settings (35).

There are limited data comparing the various AID algorithms, and none are designed for the hospital setting. Having an insulin sensitivity setting that can be changed is of great value in the hospital setting. All three commercially available AID systems allow adjustment of glucose targets. The Medtronic 670G and 770G systems facilitate two targets (120 and 150 mg/dL [exercise mode]) (37), and the Medtronic 780G (available outside the United States currently) allows three targets (100, 120, and 150 mg/dL) (35,38). The Omnipod 5 system allows five targets (110, 120, 130, 140, and 150 mg/dL), and the Tandem system allows three targets (110 mg/dL, 112.5–120 mg/dL [sleep], and 140–160 mg/dL [exercise]) (35). It is important to note that in Tandem’s sleep mode, automated correction boluses are not administered (35).

In 2020, the Continuous Glucose Monitor and Automated Insulin Dosing Systems in the Hospital Consensus Guideline Panel released recommendations regarding continuation of AID system use during hospitalization (15). These recommendations included a strong consensus that inpatient teams should plan for alternative management in case AID use is ineffective or expected to be insufficient in maintaining glycemic control (15). That alternate plan could include either transition to multiple daily insulin injections or discontinuation of the pump’s AID function in favor of its use for traditional pump therapy.

AID systems are designed to be operated in automated mode. However, the automated settings can be turned off, reducing the device capacity to that of a traditional insulin pump through “manual mode.” For systems using machine learning (i.e., Omnipod and Medtronic systems), the algorithms learn over days to weeks and are programmed to reflect the patient’s insulin requirements at steady state. Thus, patients may need to exit the automated delivery mode during acute illness if they unable to achieve glycemic targets. The consensus panelists discussed use of AID systems in manual mode but were split on a recommendation that all patients should operate their AID system only in manual mode (15).

If the automatic mode is exited, the basal suspend feature should be activated to help prevent hypoglycemia. Additionally, patients’ manually set basal rates should be evaluated to ensure accuracy and appropriate insulin delivery.

Institutional policies and guidelines should clearly outline the circumstances and qualifications for continuation of personal diabetes devices during hospitalization (12,14–17). Policies should outline the roles and responsibilities of the patient and members of the care team (Table 2). There is strong expert consensus on the need for diabetes/endocrinology team involvement during inpatient use of insulin pumps (12,14); therefore, policies should require consultation and outline the nature of expert involvement. At the authors’ institution, diabetes/endocrinology team consultation is required for all insulin pumps and AID systems; however, it is not required for the continuation of free-standing CGM systems during hospitalization because of the increasing volume of their use. Institutional policy does specify parameters for continuation of personal CGM, including the requirement for ongoing POC blood glucose monitoring to inform all insulin doses outside of the COVID-19 population.

As diabetes technology becomes increasingly popular in the outpatient setting, health systems may struggle to meet the demand if a consultation is required for each patient using an insulin pump and/or CGM. Previous reports and guidelines have not contrasted the level of expertise or involvement needed for CGM versus insulin pump therapy (14,39,40). However, endocrinology consultation is widely described in the use of hospital-furnished CGM systems during the COVID-19 pandemic, when CGM may be used nonadjunctively to directly inform treatment decisions (41–44).

Criteria for initial and continued use of personal diabetes devices in the hospital should be clearly outlined in the institutional policy (31). The policy should also include circumstances in which personal diabetes devices should be stopped during hospitalization, which include with changes in clinical condition, during procedures, and for some imaging studies.

In addition to clearly defined roles, responsibilities, and expectations, appropriate staff education regarding diabetes technology also aids in successful inpatient use of these devices. There are limited data regarding appropriate staff education for continued device use. Sweeney et al. (45) evaluated the effectiveness of insulin pump education for nurses using a 60-minute didactic session and 30-minute hands-on session for insulin pump setup and operation, which resulted in an increase in nurses’ knowledge of and confidence in insulin pump therapy. Lack of staff familiarity with these technologies and low frequency of patients wishing to use them make it important that hospital guidelines and policies are clearly articulated and easily accessible.

Although there are no reports of staff education on personal CGM use in the inpatient setting, recent COVID-19 pandemic–era reports have generated valuable information. At the authors’ institution, a scaffolded educational approach was implemented, in which the inpatient diabetes team educated nursing leaders (e.g., nurse managers and clinical nurse specialists) on CGM sensor insertion and setup. Nurse leaders then provided just-in-time training to bedside staff nurses (42). CGM training is now provided to all medical intensive care nurses as part of their annual nursing competencies. Baker et al. (46) used diabetes educators and pharmacists to provide training to nursing staff through frequent ad hoc sessions, followed by group education sessions. A general trend emphasizes in-person education paired with just-in-time training, supported by reference documents that can be kept at the bedside (42,46).

Insulin pump settings should be included in patients’ electronic health record (EHR) and pharmacy medication profile (10). This documentation can be facilitated through the creation of insulin pump order sets in the EHR system. The occurrence of an insulin bolus, along with the specific units administered for carbohydrate intake or correction dosing for hyperglycemia, should be entered into the EHR by the staff nurse. The presence of an insulin pump or CGM system, along with the condition of the pump and/or CGM site, should be documented in the EHR initially and during every shift by nursing staff (31). Documentation of a patient’s insulin pump and/or CGM use can be facilitated with the addition of a diabetes device–specific lines/drains/airways (LAD) section. Many EHR systems allow specific pump and/or CGM site locations to be marked on a mannequin. Nurses can document the date and time of site changes, along with the new location, in the LAD section. The International Classification of Disease, 10th revision, diagnostic code for “insulin pump status” (Z96.41) can be entered by the diabetes/endocrinology team or primary team. Although not billable, the code can help to identify the patient as an insulin pump user.

Clinical nonadjunctive use of CGM during the COVID-19 pandemic spurred more extensive reporting of EHR documentation solutions (44,47). Davis et al. (44) described the creation of a dedicated CGM flowsheet in the Epic EHR system that facilitated both CGM documentation and validation with a built-in calculator for POC blood glucose and CGM concordance. More advanced strategies and expert consensus on essential data elements are needed for future EHR integration of CGM in the inpatient setting.

People with diabetes are increasingly turning to technology for the self-management of their condition. As a result, the use of outpatient personal diabetes devices among patients admitted to the hospital is likely to expand moving forward. Personal diabetes devices such as CGM systems, insulin pumps, and AID systems are not FDA-approved for inpatient use; however, patients understandably may want to continue wearing these devices during hospitalization.

There is strong expert consensus that diabetes technology use can be safely and effectively continued during hospitalization for noncritically ill patients who retain mental capacity and possess sufficient self-management knowledge, with the involvement of diabetes/endocrinology teams. The degree of support required will likely vary based on the type of device used. Additionally, the mechanism of insulin delivery and available features vary significantly among the three commercially available AID systems, and these unique features should be considered when determining whether to allow use of AID during hospitalization. Data on the continuation of personal diabetes devices during hospitalization are limited. There is considerable need for further research to inform the successful inpatient use of outpatient diabetes devices.

K.M.D. and E.R.F. have received fees for consulting and speaking and research funding from Dexcom. No other potential conflicts of interest relevant to this article were reported.

All of the authors designed the review and participated in drafting the manuscript. E.R.F. is the guarantor of this work and, as such, takes responsibility for the integrity and accuracy of the article.