PROJECT SUMMARY More than 8 million people with diabetes are hospitalized each year in the United States. Hospitalized patients with uncontrolled diabetes have a disproportionate risk for poor hospital outcomes. Current strategies using multiple daily injections (MDI) of insulin often fail to improve glycemic control. Recent trials have shown that adding continuous glucose monitoring (CGM) to MDI may have a positive but modest impact on inpatient glycemic control. While high quality European studies from a single research group have shown improvement in glycemic control with automated insulin delivery (AID), several questions remain about reproducibility, sustainability, dissemination, and scalability. First, the use of expensive insulin pumps with tubing is not ideal for the hospital (potential infection transmission if used in multiple patients, reduced mobility, and cost concerns); second, the role of hospital staff in the use of AID has not been defined; third, methods for confirmation of CGM accuracy and electronic health records (EHR) documentation are not established; fourth, remote insulin delivery and remote AID and glucose monitoring has not been tested in the hospital (relevant for patients under isolation precautions); fifth, there is a need for regimens for heterogenous populations usually excluded from clinical trials (e.g. type 1 diabetes, steroid-induced diabetes). The Automated Insulin Delivery in Inpatients with Dysglycemia (AIDING) trial, informed by a recently completed feasibility pilot study, will address these questions in a multicenter randomized clinical trial (Aim 1). A heterogenous adult population with diabetes requiring insulin therapy will be randomly assigned to AID with remote real-time CGM or to MDI plus CGM. Patients in the intervention arm will wear a disposable single-user patch pump capable of remote insulin delivery. Sensor glucose values will be documented in the EHR with periodic confirmation of accuracy and will be monitored remotely by nursing staff and the study team to detect potentially dangerous hypoglycemic or hyperglycemic episodes. The primary endpoint is time spent in target glucose range (70-180 mg/dL). Safety will be assessed by time spent in hypoglycemia (<54mg/dL). In Aim 2, we will conduct a mixed-methods evaluation (surveys and semi-structured interviews) to examine additional infrastructure needs and factors that influence the adoption of inpatient AID among end-users (i.e., study participants, nursing staff) and relevant hospital staff stakeholders. This approach will facilitate the creation of standardized care and implementation algorithms to facilitate sustainability and scalability of inpatient technology use to other health systems and at- risk populations.