This randomized controlled trial will test the efficacy and safety of automated insulin delivery (AID) in hospitalized patients with diabetes (type 1 or type 2) requiring insulin therapy who are admitted to general medical/surgical floors.

The main objectives of this study are:

• To test the efficacy and safety of AID versus multiple daily insulin injections (MDI) + CGM in the inpatient setting
• To determine differences in CGM-derived metrics between AID and MDI plus CGM in the hospital and explore differences in treatment effect according to individual characteristics.

Participants will be:

• Randomized to AID + remote CGM (intervention) or multiple daily insulin injections (MDI) + CGM (control group)
• Followed for a total of 10 days or until hospital discharge (if less than 10 days).

Annually, over 8 million people in the United States are hospitalized due to diabetes. Among these patients, those with uncontrolled diabetes are at a significantly heightened risk for poor hospital outcomes. However, achieving glycemic targets within hospital settings proves challenging, as doing so often increases the risk of iatrogenic hypoglycemia. This is exacerbated by the variable insulin therapy requirements that arise during acute illnesses, such as unpredictable nutrition intake, illness severity, and the effects of various medications. These fluctuations pose a substantial challenge for healthcare providers tasked with caring for patients with diabetes.

The usage of diabetes technology may offer an opportunity to improve inpatient diabetes care, but best practices are not yet defined. The COVID-19 pandemic has highlighted how accelerated use of technologies (e.g., telemedicine, e-consults, and remote monitoring) helps healthcare systems adapt care delivery while minimizing exposure risk. Following the non-objection by the US Food and Drug Administration (FDA) to the use of CGM in the hospital, the investigators and others implemented the usage of remote real-time CGM to treat patients with COVID-19. Initial clinical trials using remote real-time CGM have shown modest improvements in hypoglycemia detection and prevention and minimal or no increases in time spent in the target range (TIR) in non-ICU patients.

The use of AID with remote insulin delivery and remote glucose monitoring is novel for the inpatient setting. The use of AID allows for 288 automated insulin dosing alterations per day, which is infeasible for hospital staff and could help proactively compensate for the multitude of factors affecting insulin requirements in the hospital. Preliminary data from AID trials with a single European AID system (using an insulin pump with tubing) have shown improvements in glycemic control in diverse populations without increasing the risk of hypoglycemia. However, these trials have limited involvement of patient care teams and the feasibility of hospital implementation and adoption is unknown.

Using single-use insulin patch-pumps (without tubing) may offer a unique opportunity for hospital use of AID; however, no randomized controlled data on the use of AID in the hospital is available in the US. Results from our recently completed pilot study show that using AID with remote CGM is feasible in the hospital and appears to be associated with good glycemic control.

In addition to improving glycemic control, this approach will:

• Offer a unique treatment option for patients typically excluded from inpatient clinical trials, including patients with type 1 diabetes, steroids, medical nutrition therapy, or those under isolation precautions.
• Use superficial wearable medical devices to alleviate inconveniences and discomforts to patients associated with current standard-of-care insulin therapy. The combination of a CGM and a patch-based insulin pump can greatly reduce the need for recurrent "sticks" and "pricks." They may thereby improve the quality of experience for patients admitted to the hospital.
• Introduce remote insulin delivery and remote glucose monitoring, offering an unprecedented opportunity for effective diabetes care for people with highly contagious diseases or while interacting with patients during emergency conditions.
• Set the stage for subsequent inpatient diabetes technology implementation efforts. The investigator's approach has multiple layers of safety to ensure improved glycemic control without increasing the risk of hypoglycemia, including a) nursing staff training for rapid response, b) EHR documentation and validation of sensor glucose values to confirm accuracy, c) remote monitoring with alarms (telemetry and inpatient treatment team).

Inclusion Criteria:

• Any person ≥18 years of age with diabetes mellitus (except cystic fibrosis- and pregnancy-related) admitted to general (non-ICU) medical-surgical hospital services which require inpatient insulin therapy (i.e.,TID or T2D with ≥2 glucose values ≥180mg/dl)

Exclusion Criteria:

• Patients admitted to ICU
• Patients anticipated to require less than 48 hours admission.
• Current evidence of hyperglycemic crises (diabetic ketoacidosis or hyperosmolar hyperglycemic state)
• Severe anemia with hemoglobin <7 g/dL
• Evidence of hemodynamic instability
• Hypoxia (SpO2 <92% on supplemental oxygen)
• Pre-admission or inpatient total-daily insulin dose >150 units daily
• T2D patients on sliding scale insulin therapy alone (no scheduled basal or bolus insulin) and with glucose levels below 180 mg/dl
• Patients without diabetes with stress hyperglycemia (not related to steroids or medical nutrition therapy) and with HbA1c <6.5%
• Patients on AID as outpatient
• Patients who previously participated in AIDING feasibility trial or this RCT
• Patients with a condition impeding their ability to consent or answer questionnaires or notify staff of symptoms.
• Patients who are pregnant time of enrollment
• Patients who are unable or unwilling to use rapid-acting insulin analogs (Humalog, Admelog, or Novolog) during the study
• Use of hydroxyurea, high dose of acetaminophen (>4 grams/day), or high dose ascorbic acid