Hyperglycemia alters immune response by inhibiting chemotaxis and phagocytosis. It affects the bactericidal ability of immune cells by decreasing the production of superoxide radicals [17]. Hyperglycemia also induces osmotic diuresis, endothelial injury, and mitochondrial dysfunction, which can lead to shock and multiple organ failure in hospitalized patients [18]. It has been shown that hyperglycemia in hospitalized patients is associated with an increased risk of complications, mortality, a higher rate of admission to ICUs, and a higher need to transition to rehabilitation facilities after discharge [19,20]. It has been reported that the risk of postoperative infections in general surgery patients increases by 30% for every 40 mg/dL (2.2 mmol/L) rise in blood glucose over normoglycemia (blood glucose of 110 mg/dL or 6.1 mmol/L) [21]. In a case–control study of 2236 surgical patients, patients with glucose levels of 110 to 200 mg/dL (6.1–11.1 mmol/L) and glucose levels above 200 mg/dL (>11.1 mmol/L) had a 1.7- and 2.1-fold increase in mortality (respectively) compared to patients with blood glucose less than 110 mg/dL (<6.1 mmol/L) [22]. In a prospective study of 2471 patients with community-acquired pneumonia, patients with admission blood glucose above 198 mg/dL (11 mmol/L) had a greater mortality and complications compared to patients with blood glucose less than 198 mg/dL (<11 mmol/L) (13% vs. 9%, p = 0.03) [23]. In this study, the risk of complications in these patients increased by 3% for each 18 mg/dL (1 mmol/L) rise in admission blood glucose above 110 mg/dL (6.1 mmol/L) [23].

Many other studies have also suggested that the early intervention and treatment of inpatient hyperglycemia can reduce the risks of hospital-acquired infections and decrease the length of hospital stay and mortality [24,25].

A recent meta-analysis on 5053 surgical patients showed tight blood glucose control (<110–120 mg/dL or 6.1–6.7 mmol/L) as compared to conventional blood glucose control (<200 mg/dL or <11.1 mmol/L) was associated with a lower risk of postoperative infections (9.4% vs. 15.8%, p < 0.001), lower wound infection rates (4.6% vs. 7.2%, p = 0.015), and higher post-operation hypoglycemia rates (22.3% vs. 11%, p < 0.001). However, the short term mortality was not significantly different in both groups. Therefore, it was suggested that the higher risk of hypoglycemic events in tight blood glucose control group had contributed to the mortality rate in this group [26]. In another study in patients in the ICU, patients with blood glucose levels more than 200 mg/dL (>11.1 mmol/L) were found to have higher mortality comparing to patients with blood glucose levels less than 200 mg/dL (<11.1 mmol/L) (5.0% vs. 1.8%, p < 0.001) [27].

More recent studies during the COVID-19 pandemic also showed that hyperglycemia was commonly seen in hospitalized patients with COVID-19 in ICUs (65.5% of patients), which reflected the severity of the condition and was also related to the treatment with steroids and was associated with poorer outcomes (HR 3.535, 95% confidence interval [CI] 1.338–9.338) [28,29].

Hypoglycemia in hospitalized patients is categorized in three levels by the American Diabetes Association (ADA) Standards of Care. Level 1 is defined as blood glucose between 54 mg/dL (3 mmol/L) and 70 mg/dL (3.9 mmol/L); level 2 is a blood glucose < 54 mg/dL (<3 mmol/L), which is the threshold for neuroglycopenic symptoms; and level 3 is any blood glucose level with altered mental and/or physical status that requires assistance from others for recovery [30].

It has been shown that hypoglycemia is associated with increased levels of proinflammatory cytokines (TNF-α, IL-1β, IL-6, and IL-8), markers of lipid peroxidation, and oxidative stress and these changes can contribute to adverse cardiovascular events associated with hypoglycemia [31]. It has been reported that hypoglycemia in hospitalized patients is associated with poor cardiovascular outcomes including prolonged QT interval, angina, ischemic changes, arrhythmia, and sudden death [32,33].

The incidence of severe hypoglycemia (blood glucose < 40 mg/dL or 2.2 mol/L) in ICU patients has been reported in 28% of patients, and in non-ICU settings, as high as 33% of patients [34]. The risk of hypoglycemic events increases in hospitalized patients with older age, the use of oral hypoglycemic medications, the presence of severe illness, and unexpected changes in nutritional intake and steroid tapering [35].

In a cohort of 5365 patients in a medical–surgical ICU, the odds ratio for mortality associated with one or more episode of hypoglycemia was 2.28 (1.41–3.7, p = 0.0008) [36]. A 2019 systemic review suggested that hospitalized patients on general medical–surgical wards with an episode of blood glucose less than 72 mg/dL (4 mmol/L) had higher in-hospital mortality than patients without hypoglycemia (95% confidence interval [CI]: 1.64–2.67) [37].

Based on compelling evidence, it has been shown that glycemic excursions including hyperglycemia and severe hypoglycemia are associated with increased adverse outcomes in hospitalized patients. Therefore, many clinical investigators have tried to define optimal glucose targets for glycemic control in hospitalized patients.

The Leuven Surgical ICU study was the pioneer of trials that promoted intensive glucose control in ICU settings. In this study, 1548 patients in surgical ICUs were randomized to conventional blood glucose control with a target of 180–200 mg/dL (10–11.1 mmol/L) or the intensive therapy group with a target of 80–110 mg/dL (4.4–6.1 mmol/L). A total of 63% of these patinets were cardiac surgery patients and 13% had pre-existing diabetes, and almost all patients received early parenteral nutrition after surgery. The patients in the intensive glycemic targets group compared to patients with the conventional blood glucose targets had significantly less bacteremia, lower ventilator dependency, and shorter ICU stays, and had an overall 34% reduction in mortality [38]. However, subsequent studies did not confirm these findings and the majority of them reported that lower glucose targets were associated with more hypoglycemic events and even increased mortality in patients in ICUs [39,40].

In the Glucontrol trial, a seven-country randomized clinical trial (RCT), 1101 medical and surgical patients in ICU were randomized to tight glycemic targets of 80–110 mg/dL (4.4–6.1 mmol/L) vs. conventional glycemic control of 140–180 mg/dL (7.8–10 mmol/L). No statistically significant difference in mortality was detected in either group (15.3% vs. 17.2%). However, the risk of hypoglycemia was increased in the intensive tight glycemic group (8.7% vs. 2.7% p < 0.0001) [39]. In the Efficacy of Volume Substitution and Insulin Therapy in Sepsis (VISEP) trial from Germany, 600 patients with sepsis in ICUs were randomized to conventional glucose control of 180–200 mg/dL (10–11.5 mmol/L) vs. intensive glucose control with a target of 80–110 mg/dL (4–6.1 mmol/L). This trial was stopped early as no significant difference between 28- and 90-day mortality rates was detected in either group (21.6% vs. 21.9% and 29.5% vs. 32.8%, respectively); however, those in the intensive group had a significantly higher rate of hypoglycemic events (12.1% vs. 2.1%) [41].

In the Normoglycemia in Intensive Care Evaluation and Surviving Using Glucose Algorithm Regulation (NICE-SUGAR) trial, 6104 patients in ICU were randomized to the conventional glycemic control with a target of blood glucose < 180 mg/dL (10 mmol/L) or intensive glucose control with a target of 81–108 mg/dL (4.4–5.7 mmol/L) [42]. Data from the NICE-SUGAR trial reported that hospital mortality was not significantly different in both groups but the mortality rate at 90 days was significantly higher in the intensive group (24.9% vs. 27.5% p = 0.02), with a higher incidence of hypoglycemia events (0.5% vs. 6.8%) in the intensive group [42]. A subsequent analysis of data by the NICE-SUGAR investigators also reported that patients with hypoglycemia had a two-fold increase in mortality compared to patients without hypoglycemic events [43].

Additionally, many studies have documented the significant negative impact of hyperglycemia in the hospital outcomes of cardiac surgery patients. A retrospective study of 409 cardiac surgery patients showed that every 20 points above a blood glucose of 100 mg/dL (5.6 mmol/L) is associated with a 30% increase in adverse outcomes, including pulmonary and renal complications and mortality [44]. A meta-analysis of five randomized clinical trials (RCTs) in 706 cardiac surgery patients showed that intensive glycemic control was associated with a decreased rate of infections (risk ratio = 0.50; 95% confidence Interval [CI] = 0.29–0.84; p = 0.009) but was not associated with a decreased mortality (95% confidence interval [CI] = −0.01 to 0.03; p = 0.25) [45].

In the GLUCO-CABG trial, 302 post-CABG patients with stress hyperglycemia with diabetes and without diabetes were assigned to computerized intensive blood glucose control with a target of 100–140 mg/dL (5.6–7.8 mmol/L) and conservative blood glucose control with a target of 141–180 mg/dL (7.8–10 mmol/L) [46]. In this trial, the intensive group had a 20% reduction (nonsignificant) in perioperative complications (42% vs. 52%; p = 0.08). [46] In addition, no significant difference in the rate of complications and mortality among patients with diabetes in both groups was detected (49.3% vs. 45.8%; p = 0.65) [46].However, in patients with stress hyperglycemia and without pre-existing diabetes, the rate of complications (including sternal wound infections, bacteremia, respiratory failure, pneumonia, acute kidney injury, stroke, cardiac arrhythmia, and heart failure) was significantly lower in the intensive group than in the conservative glucose control group (35% vs. 58%; p = 0.006) [46]. Hospitalization costs were also significantly lower in patients with stress hyperglycemia in the intensive group compared to the conservative glucose control group (USD 36.681 vs. USD 40,913; p = 0.04). The rate of mild hypoglycemia was significantly higher in the intensive group (9% vs. 3%; p = 0.09); however, it was still lower than the reported hypoglycemic events in patients in the ICU. The lower rate of mild hypoglycemia events in this study was attributed to the use of computerized insulin infusion protocols. In this trial, no severe hypoglycemia less than 40 mg/dL (2.2 mmol/L) was reported in patients in the intensive and conventional glycemic groups [46]. These findings suggest that in selected patients in ICUs, such as cardiac surgery patients, lower glycemic targets in experienced centers while avoiding severe hypoglycemia may be beneficial.

Based on a large number of observational studies and randomized trials including the above-mentioned studies, many professional societies including ADA (the American Diabetes Association), AACE (the American Association of Clinical Endocrinologists), and the Endocrine Society recommend glycemic targets of 140 to 180 mg/dL (7.8–10 mmol/L) for the majority of patients in ICUs and a lower glucose target of 110–140 mg/dL (6.1–7.8 mmol/L) for selected patients in ICUs (including post-cardiac-surgery patients in medical centers with expertise) [8,19,30]. Glucose levels above 180 mg/dL (10 mmol/L) and lower than 110 mg/dL (6.1 mmol/L) are not recommended in ICU patients [19,30]. Additionally, the Society of Critical Care Medicine recommends blood glucose targets of lower than 180 mg/dL (10 mmol/L) in patients in ICUs and recommends initiation of treatment when glucose is above 150 mg/dL (8.3 mmol/L) [47].

Intravenous (IV) insulin infusion is the preferred method for achieving blood glucose targets in ICU patients as it offers flexibility in the care of critically ill patients in ICUs with their need for treatment adjustments in response to frequent changes in their blood glucose and nutritional intake [48]. There is no ideal protocol or clear evidence that supports the benefit of one protocol over any others [49].

A proper insulin infusion protocol offers the ability to adjust the infusion rate based on the current and previous glucose values and the rate of glucose changes and the ability to closely monitor blood glucose changes with hourly glucose measurements and adjustments [50]. In some institutions, software-based or computerized algorithms have been used for IV insulin infusion protocols. Proportional–integral–derivative (PID) models are mostly used in software-based protocols. In these models, previous glucose levels are used to titrate the insulin infusion rate by using a dynamic multiplier responsive to insulin sensitivity and changes in glucose levels for a given insulin dose [51]. Recent randomized clinical trials (RCTs) and retrospective cohorts have reported more rapid and tighter glycemic control and lower glycemic variability with computerized algorithms than the standard paper protocol [52,53]. In a recent study, 61 patients in ICUs in a computerized protocol group were compared to 51 patients in ICUs with similar demographics that were assigned to the standard insulin protocol, and it was reported that the computerized group had a greater percentage of glucose measurements that were within the target range compared to the standard insulin protocol group (68.4% vs. 36.5%, p < 0.001) [53]. The computerized group also had a shorter time-to-target [median (interquartile range) 5 h (3–8 h) vs. 7 h (4–10 h); p = 0.02] and lower severe hypoglycemic events (26 vs. 6, p < 0.0001) [53]. However, these findings were not confirmed by other studies, including a multicenter study which evaluated a computerized insulin protocol and standard insulin infusion protocol in 1300 patients in 34 French ICUs [54]. This study reported no statistically significant difference between the computerized vs. standard insulin infusion protocols, and reported more hypoglycemic events in the computerized insulin protocol [54].

For noncritically ill medical and surgical patients, ADA and AACE recommend a target of premeal blood glucose lower than 140 mg/dL (7.8 mmol/L) and random blood glucose lower than 180 mg/dL (10 mmol/L) in non-ICU settings [19,30]. Subcutaneous insulin is the treatment of choice for noncritically ill patients. However, the use of sliding-scale insulin alone for hospitalized patients with diabetes is not recommended [19,30]. For hospitalized patients with type 1 diabetes, the use of basal insulin plus short- or rapid-acting insulin for meal coverage is recommended [30]. Many recent studies have reported that the use of subcutaneous basal insulin plus short- or rapid-acting insulin prior to meals (basal/bolus) is the most appropriate and safe treatment for the management of hyperglycemia in hospitalized patients with type 2 diabetes [19,30,55]. In a prospective randomized multicenter trial, the Randomized Study of Basal Bolus Insulin Therapy (RABBIT-2) Medicine, 130 patients with type 2 diabetes were assigned randomly to a basal/bolus insulin regimen vs. sliding-scale alone. This study showed that 66% of patients in the basal/bolus regimen group achieved blood glucose lower than 140 mg/dL (7.8 mmol/L) compared to 38% of patients in the sliding-scale group (p < 0.05) [56]. In the RABBIT-2 trial, the incidence of hypoglycemia was the same in both groups (<5% of patients) [56]. In another similar study, a randomized multicenter trial (Randomized Study of Basal Bolus Insulin Therapy (RABBIT-2) Surgery), 212 surgical patients with type 2 diabetes were assigned randomly to the sliding-scale insulin regimen vs. basal/bolus insulin regimen. In this study, blood glucose lower than 140 mg/dL (7.8 mmol/L) was achieved in 55% of patients in the basal/bolus insulin group vs. 31% of patients in the sliding-scale group (p < 0.001) [57]. Additionally, the rate of complications such as wound infection, pneumonia, acute kidney injury, and acute respiratory failure was reduced in the basal/bolus insulin regimen group vs. the sliding-scale group (8.6% and 24.3%; odds ratio 3.39 (95% confidence interval [CI]: 1.50–7.65); p = 0.003) [57]. However, hypoglycemia with blood glucose lower than 70 mg/dL (3.9 mmol/L) was seen in 23.1% of patients in the basal/bolus group vs. 4.7% of patients in the sliding-scale insulin group (p < 0.001) [57]. Due to the higher risk of hypoglycemia in the basal/bolus insulin group in surgical patients, a subsequent study was performed to evaluate the treatment protocols of basal insulin plus a lower insulin scale defined as a correction scale (basal plus) vs. basal/bolus insulin vs. sliding-scale insulin regimens in insulin-naïve hospitalized patients on only oral agents at home or in patients with very minimal insulin usage at home (<0.4 units/kg/day) and in patients with poor oral intake in the hospital [58]. In this study, 375 patients with diabetes were randomly assigned to basal plus (basal insulin plus correction scale) vs. basal/bolus vs. sliding scale. Both basal/bolus and basal plus groups did better than the sliding-scale group in achieving the target of blood glucose of lower than 140 mg/dL (7.8 mmol/L) (p = 0.04) [58]. The basal/bolus and basal plus groups had less treatment failure than the sliding-scale group (0% vs. 2% vs. 19%, respectively; p < 0.001) [58]. The incidence of hypoglycemia with blood glucose lower than 70 mg/dL (3.9 mmol/L) was higher in both the basal/bolus and basal plus groups (16% vs. 8%, p = 0.48) than in the sliding-scale group (3%) (p <0.05) [58]. However, the incidence of severe hypoglycemia with blood glucose lower than 40 mg/dL (2.2 mmol/L) was not significantly different among the three treatment groups (1% vs. 1% vs. 0%, respectively; p = 0.76) [58]. Therefore, for surgical patients with poor nutritional intake or at higher risk of hypoglycemia and patients on low doses of insulin at home, it is recommended to use the basal insulin plus correction scale with short-acting insulin rather than using meal boluses or the sliding-scale alone.

The recommended dose of insulin for most people with type 2 diabetes admitted to the hospital is 0.3–0.5 units/kg/day. However, the starting dose in elderly patients and patients with impaired renal function should be lower to decrease the risk of hypoglycemia in these high-risk groups [59].

The use of noninsulin therapies in hospitalized patients with type 2 diabetes is not recommended [30], ADA and AACE recommend against using metformin in hospitalized patients due to the presence of dehydration and acute kidney injury in many patients [19,30]. Sulfonylurea can increase the risk of severe hypoglycemia and is not recommended for inpatient settings [30]. Glucagon-like peptide-1 (GLP-1) agonists are also not well tolerated in hospitalized patients due to gastrointestinal side effects including nausea and vomiting [30]. However, recent studies have shown that Dipeptydil Peptidase IV (DPP-IV) inhibitors alone or with basal insulin may be a safe and effective choice in noncritically ill patients [30]. However; among DPP-IV inhibitors, saxagliptin and alogliptin are contraindicated in patients with congestive heart failure [30,60,61]. Sodium Glucose cotransporter-2 (SGLT-2) inhibitors are not recommended by the FDA for inpatient glycemic control due to the increased risk of euglycemic diabetic ketoacidosis (e-DKA), and they should be held for 3–4 days prior to any elective surgery [30].

It has been recommended that all surgical patients, patients with diabetes, and hospitalized patients with stress hyperglycemia with blood glucose > 140 mg/dL (7.8 mmol/L) have a nutrition assessment in the first 24 h of hospital admission [19,30]. The daily energy intake requirement for patients with diabetes is usually met with 25–35 calories/kg/day, while critically ill patients may require less with the target of 15–25 calories/kg/day [62]. Enteral nutrition is the second-best option after oral nutrition as it is associated with a lower risk of complications, a lower risk of gastric mucosa atrophy, and a lower risk of infection and thrombosis compared to parenteral nutrition [63,64]. Standard enteral formulas provide lower amounts of lipids (30% of total calories) combined with higher carbohydrate content (55–66% of total calories); however, diabetic specific formulas (DSFs) have replaced some carbohydrates with monosaturated fatty acids (up to 35% of total calories), fibers (10–15% of total calories), and fructose (up to 30% of total calories) [64,65]. It has been reported that the postprandial spike in blood glucose was reduced by 18–29 mg/dL (0.8–1.7 mmol/L) with the use of DSF in noncritically ill patients, with no significant increase in LDL cholesterol or increase in the risk of lactic acidosis [65].

For continuous enteral feeding, the use of basal insulin with short- or rapid-acting insulin boluses per sliding scale every 4–6 h is recommended [30]. Basal insulin is calculated based on patients’ total daily dose plus additional insulin for enteral feeding which is calculated as 1 unit of insulin for each 10–15 g of carbohydrates in the enteral feeding formula. If enteral feeding is interrupted, the immediate use of IV dextrose (D10) at 50 mL/h is recommended to avoid hypoglycemia [30]. For bolus enteral feedings, the use of 1 unit of regular insulin for each 10–15 g of carbohydrates in the enteral feeding formula plus correctional insulin before each feeding is recommended. For nocturnal tube feeding, the use of insulin NPH given at the start of feeding is appropriate [30,62,63]. In critically ill patients, the use of parenteral nutrition is recommended with intravenous insulin infusion for the management of hyperglycemia [30,62,66]. To reduce hyperglycemia, glucose content in parenteral nutrition should be limited to 150–200 g/day. The addition of regular insulin at 1 unit for each 10 g of dextrose in a parenteral nutrition bag is also recommended by the ADA for patients with diabetes to avoid hypoglycemia with an interruption of parenteral nutrition [30].