Glycemic Management in Insulin-Naive Patients in the Inpatient Setting

Abstract

Objective

The ideal inpatient insulin regimen efficiently attains the target blood glucose range, effectively treats hyperglycemia, and minimizes the risk of hypoglycemia. The objective of this study was to compare glycemic targets achieved by using correctional monotherapy (CM) and basal-bolus therapy (BBT) in insulin-naive patients in the inpatient setting to determine optimal blood glucose management for these patients.

Design

This was a retrospective observational cohort study of 792 patients with diabetes not on home insulin therapy who were admitted to an academic hospital over a 5.5-month period. The percentages of hyperglycemic and hypoglycemic values in each group were compared.

Results

Among the 3,112 measured blood glucose values obtained from 792 patients within the first 24 hours of insulin administration, 28.5% were hyperglycemic in the BBT group compared with 23.5% in the CM group. When adjusted for covariates, there was a 23% decrease in hyperglycemia in the BBT group (incidence rate ratio = 0.77, 95% CI 0.64–0.95, P = 0.006). Increases in A1C and admission blood glucose, as well as decreases in admission creatinine and inpatient steroid use, were independently associated with higher rates of hyperglycemia, adjusted for all other covariates. There was no significant difference between the groups in the rate of hypoglycemia in the first 24 hours, which was 1.9% in the BBT group and 1.4% in the CM group (P = 0.301).

Conclusion

Utilizing BBT in insulin-naive patients admitted to the hospital within the first 24 hours of insulin administration results in lower rates of hyperglycemia without higher rates of hypoglycemia when compared with CM.

Inpatient diabetes management is an increasingly important topic because of the prevalence of this common disease, the costs associated with inpatient glycemic management, and the association of abnormal blood glucose values in the inpatient setting with poor outcomes.

According to the Centers for Disease Control, 34.2 million Americans, or 10.5% of the U.S. population, have diabetes (1). Mostly because of the aging of the U.S. population and people with diabetes living longer, the percentage of patients with diabetes is only expected to increase, with projections that up to 33% of U.S. adults will have diabetes by 2050 if recent increases in diabetes incidence continue (2). In 2016 alone, there were >7.8 million hospital stays for patients with diabetes either as a principal or secondary diagnosis for hospitalization, and patients with a diagnosis of diabetes admitted to the hospital have longer lengths of stay than those without diabetes (1).

The total estimated direct and indirect costs of diagnosed diabetes care in the United States in 2017 was $327 billion, of which 73% were direct health care expenses attributed to diabetes and 27% were related to work-related productivity (3). Of the 162 million hospital inpatient days in the United States in 2017, an estimated 40.3 million days were incurred by people with diabetes, of which 22.6 million days were attributed to diabetes.

Hyperglycemia is not a physiological or benign condition but rather a marker of increased inpatient mortality, morbidity, hospital costs, hospital length of stay, and readmission (4,5). Despite the available management options for hyperglycemia, glycemic management in hospitalized patients remains suboptimal. Both spontaneous and iatrogenic hypoglycemia are limiting factors to normalizing blood glucose levels in the inpatient setting, and hypoglycemia is an independent predictor of poor outcomes, including longer hospital stays, higher risk of inpatient mortality, increased hospital costs, and increased likelihood of discharge to a skilled nursing facility (6,7).

In this study, we took a retrospective approach to compare the difference in glycemia using two different insulin regimens for insulin-naive patients admitted to the hospital. We aimed to determine whether there was improved glycemia with the use of a basal-bolus therapy (BBT) compared with a correctional monotherapy (CM) insulin regimen in these hospitalized patients.

Research Design and Methods

This was a retrospective, single-center, observational cohort study of 792 patients ≥18 years of age with a known history of type 2 diabetes or with type 2 diabetes diagnosed on admission. Patients were admitted to an academic institution between 15 October 2019 and 31 March 2020. Further inclusion criteria included outpatient diabetes management with either lifestyle modifications alone or any combination of noninsulin antihyperglycemic medications. Exclusion criteria included type 1 diabetes, home insulin use, diabetic ketoacidosis, hyperglycemic hyperosmolar state, and pregnancy. For patients who were hospitalized multiple times during the study period, only the initial hospitalization was included.

This study was approved by the institutional review board and granted exemption from informed consent requirements because of its retrospective design. The work was carried out in accordance with the Declaration of Helsinki for experiments involving humans.

Patient demographics and clinical characteristics were collected, including A1C, home antihyperglycemic medications, inpatient oral corticosteroid use with no dose threshold, inpatient treatment plan, admission blood glucose, admission creatinine, and blood glucose values within 72 hours of admission from the onset of initial insulin administration.

In our institution, glargine is used as basal insulin and lispro is used as bolus and correctional insulin. Glycemia is managed by the primary care team on both medicine and nonmedicine services unless the endocrinology team is consulted to assist with glycemic management. Providers are given clinical decision support within the electronic medical record system when initiating insulin therapy, but ultimately can choose which insulin regimen should be initiated for any given patient.

Medical records were analyzed for initiation of a CM or BBT insulin regimen over the first 72 hours of hospital admission from the onset of insulin administration. Patients were included in the CM group if they were started on correctional short-acting insulin alone, and patients were included in the BBT group if they were started on both basal and bolus insulin or either one, in addition to correctional insulin. In instances in which patients were initially started on one regimen and then subsequently transitioned to the other, only glucose values measured before the time of transition were recorded. Dose adjustments were made, with the degree of escalation or de-escalation determined by the primary care team. Patients who had dose adjustments but continued the same type of regimen were kept in their original group.

Glucose values from within 72 hours of admission from the onset of initial insulin administration were reviewed. Glycemic assessment was based on glucose readings, with glycemic targets for patients with diabetes defined according to the American Diabetes Association (ADA) recommendations (8). Per our institutional classification, hypoglycemia was defined as glucose <70 mg/dL, severe hypoglycemia as <40 mg/dL, and hyperglycemia as >180 mg/dL. Initial glucose values were obtained from serum venous glucose sources, with successive values taken from capillary point-of-care (POC) glucose monitoring tests. Hypoglycemia during hospital admission initiated a hypoglycemia protocol, through which more frequent POC glucose measurements were obtained. For the purposes of our study, only the initial hypoglycemic value and the single subsequent POC glucose value obtained 15 minutes after treatment were included.

Demographic and clinical characteristics were summarized and presented by group using the median (interquartile range [IQR]) or frequency (percentage), as appropriate. Continuous variables were assessed for normality using the Kolmogorov-Smirnov test, histogram, and Q-Q plot. Baseline characteristics were compared between groups using the Wilcoxon rank-sum test for continuous variables and χ² or Fisher exact test for categorical variables. Because the numbers of hyperglycemia values were over-dispersed count data, univariate and multivariable negative binomial regression models were used to analyze these data using the total number of glucose values as the offset variable. Deviance statistics and Akaike information criterion were used to assess the goodness of fit for these models. Percentages of hyperglycemia and hypoglycemia were computed using the algorithm total number of hyperglycemia or hypoglycemic events divided by the total number of blood glucose values within the specific category of each relevant variable. Assuming each hyperglycemia or hypoglycemic event was independent, a χ² or Fisher exact test was used to compare these percentages between treatment groups. SAS, v. 9.4, statistical software was used to conduct all analyses, and statistical significance was assumed at P <0.05.

Results

A total of 792 patients meeting inclusion and exclusion criteria were admitted during the study period. Of these, 227 patients were initiated on BBT and 565 were initiated on CM. Baseline demographics and clinical characteristics are depicted in Table 1.

Table 1.

Demographic and Clinical Characteristics

	CM (n = 565)	BBT (n = 227)	Overall (N = 792)	P *
Demographics
Age, years	73.0 (63.0–82.0)	72.0 (63.0–81.0)	73.0 (63.0–81.0)	0.368
BMI, kg/m2	28.4 (25.0–33.8)	29.4 (25.4–33.9)	28.8 (25.0–33.8)	0.275
Female sex	265 (46.9)	99 (43.6)	364 (46.0)	0.401
Race/ethnicity  Caucasian  African American  Asian  Hispanic  Other	379 (67.1) 90 (15.9) 20 (3.5) 15 (2.7) 61 (10.8)	161 (70.9) 35 (15.4) 3 (1.3) 8 (3.5) 20 (8.8)	540 (68.2) 125 (15.8) 23 (2.9) 23 (2.9) 81 (10.2)	0.384
Clinical characteristics
A1C, %	6.8 (6.2–7.5)	7.2 (6.4–8.2)	6.9 (6.2–7.7)	<0.001
Admission blood glucose, mg/dL	147.0 (118.0–190.0)	174.0 (132.0–227.0)	152.0 (121.0–201.0)	<0.001
Admission creatinine, mg/dL	1.0 (0.8–1.4)	1.1 (0.8–1.3)	1.0 (0.8–1.3)	0.712
Steroid use	113 (20.0)	57 (25.1)	170 (21.5)	0.113
Number of outpatient antihyperglycemic medications  0  1  2  ≥3	96 (17.0) 310 (54.9) 118 (20.9) 41 (7.3)	24 (10.6) 117 (51.5) 62 (27.3) 24 (10.6)	120 (15.2) 427 (53.9) 180 (22.7) 65 (8.2)	0.02

Data are n (%) or median (IQR).

^*P values are from Wilcoxon rank-sum test for continuous variables and χ² or Fisher exact test for categorical variables.

Of the 3,112 blood glucose values that were measured within the first 24 hours of admission from the onset of insulin administration, 781 (25.1%) were hyperglycemic, 2,282 (73.3%) were euglycemic, and 49 (1.6%) were hypoglycemic. In the BBT group, 28.5% of blood glucose values were hyperglycemic compared with 23.5% in the CM group. In an adjusted negative binomial model, the BBT group had a 23% lower rate of hyperglycemia than the CM group (incidence rate ratio [IRR] 0.77, 95% CI 0.64–0.95, P = 0.006). Increases in A1C and admission blood glucose, decreases in admission creatinine, and steroid use were independently associated with higher rates of hyperglycemia when adjusted for all other covariates (Table 2). There was also no significant difference in rates of hypoglycemia between the two groups, with 1.9% of blood glucose values hypoglycemic in the BBT group compared with 1.4% in the CM group (P = 0.301).

Table 2.

Negative Binomial Models for Percentage of Hyperglycemia Within 24 Hours

Variable	Unadjusted IRR (95% CI)	P	Adjusted IRR (95% CI)	P
Treatment plan  CM  BBT	Ref 1.20 (0.98–1.47)	0.075	Ref 0.77 (0.64–0.93)	0.006
Age, for every 5-year increase	0.94 (0.90–0.97)	<0.001	1.0 (0.96–1.04)	0.934
BMI	1.01 (1.00–1.02)	0.079	1.004 (0.99–1.01)	0.537
Female sex	0.92 (0.76–1.11)	0.386	0.86 (0.72–1.03)	0.093
Ethnicity  Caucasian  African American  Asian  Hispanic  Other	Ref 1.30 (1.01–1.70) 1.10 (0.62–1.94) 1.71 (1.03–2.86) 1.10 (0.81–1.51)	0.042 0.747 0.039 0.533	Ref 1.07 (0.85–1.35) 1.18 (0.69–2.02) 0.99 (0.63–1.57) 0.95 (0.72–1.25)	0.542 0.551 0.966 0.697
A1C, for every 1% increase	1.46 (1.37–1.55)	<0.001	1.24 (1.15–1.33)	<0.001
Admission blood glucose, for every 10 mg/dL increase	1.08 (1.06–1.09)	<0.001	1.05 (1.04–1.07)	<0.001
Admission creatinine	0.86 (0.68–1.08)	0.186	0.78 (0.62–0.98)	0.031
Steroid use	1.73 (1.40–2.13)	<0.001	1.88 (1.54–2.28)	<0.001
Number of outpatient antihyperglycemic medications  0  1  2  ≥3	Ref 1.01 (0.76–1.35) 1.41 (1.03–1.93) 1.71 (1.17–2.52)	0.926 0.031 0.006	Ref 1.10 (0.85–1.41) 1.31 (0.99–1.74) 1.39 (0.99–1.94)	0.483 0.061 0.057

Bold type indicates statistical significance. Ref, reference category.

When extended to include 7,928 blood glucose values over the first 72 hours of admission from the onset of insulin administration, 2,222 (28.0%) were hyperglycemic, 5,614 (70.8%) were euglycemic, and 92 (1.2%) were hypoglycemic. In the BBT group, 26.2% of blood glucose values were hyperglycemic compared with 23.4% in the CM group. In an adjusted negative binomial model, the BBT group had a 22% lower rate of hyperglycemia than the CM group (IRR 0.78, 95% CI 0.66–0.93, P = 0.004). Increases in A1C and admission blood glucose, use of three or more outpatient antihyperglycemic medications, and steroid use were independently associated with higher rates of hyperglycemia when adjusted for all other covariates (Table 3). There was, however, a significant difference in rates of hypoglycemia between the two groups, with 1.9% of blood glucose values hypoglycemic in the BBT group compared with 0.84% in the CM group (P <0.001).

Table 3.

Negative Binomial Model for Percentage of Hyperglycemia Within 72 Hours

Variable	Unadjusted IRR (95% CI)	P	Adjusted IRR (95% CI)	P
Treatment plan  CM  BBT	Ref 1.13 (0.94–1.36)	0.187	Ref 0.78 (0.66–0.93)	0.004
Age, for every 5-year increase	0.96 (0.93–0.99)	0.014	1.02 (0.99–1.06)	0.159
BMI	1.01 (1.0–1.02)	0.167	1.0 (0.99–1.01)	0.738
Female sex	0.89 (0.76–1.06)	0.19	0.84 (0.72–0.98)	0.027
Ethnicity  Caucasian  African American  Asian  Hispanic  Other	Ref 1.22 (0.97–1.53) 1.01 (0.61–1.66) 1.82 (1.14–2.89) 1.12 (0.85–1.48)	0.09 0.974 0.012 0.405	Ref 1.01 (0.82–1.24) 0.91 (0.56–1.50) 1.32 (0.86–2.04) 1.07 (0.84–1.36)	0.95 0.717 0.203 0.587
A1C	1.46 (1.37–1.55)	<0.001	1.27 (1.19–1.36)	<0.001
Admission blood glucose, for every 10 mg/dL increase	1.07 (1.06–1.08)	<0.001	1.05 (1.03–1.06)	<0.001
Admission creatinine	0.94 (0.77–1.15)	0.556	0.91 (0.75–1.10)	0.307
Steroid use	1.51 (1.24–1.83)	<0.001	1.72 (1.43–2.06)	<0.001
Number of outpatient antihyperglycemic medications  0  1  2  ≥3	Ref 1.02 (0.80–1.30) 1.46 (1.11–1.93) 1.83 (1.30–2.59)	0.884 0.006 <0.001	Ref 1.10 (0.88–1.38) 1.29 (1.0–1.65) 1.63 (1.20–2.22)	0.395 0.05 0.002

Bold type indicates statistical significance. Ref, reference category.

Discussion

The high prevalence of diabetes and its complications make it a common comorbid condition in hospitalized patients. The inpatient setting itself provides obstacles to managing glycemia. Such obstacles include, but are not limited to, physiological responses to acute illness, inconsistent caloric intake, unanticipated changes in nutrition, changes from home medications, use of medications associated with increased insulin resistance, comorbid events such as fluctuating renal function, and limitations regarding the timing of glucose monitoring and insulin administration. The ideal inpatient insulin regimen efficiently attains the target blood glucose range, effectively treats hyperglycemia, and minimizes the risk of hypoglycemia.

Our study examines the differences in glycemic management with BBT versus CM in inpatient practice outside of a protocol-driven, randomized controlled trial (RCT) and therefore contributes new findings about glycemic management in actual practice, specifically within the first 24 and 72 hours of hospitalization. Our study demonstrates that, when adjusted for covariates, the initiation of BBT for glycemic management of insulin-naive patients in the inpatient setting within the first 24 hours of admission from the onset of insulin administration is associated with lower rates of hyperglycemia without higher rates of hypoglycemia when compared with CM. Adjusted for other covariates, the major contributing factor driving the difference in hyperglycemia was steroid use, followed by differences in A1C, admission creatinine, and admission blood glucose. When extended out to include blood glucose values throughout the first 72 hours, the rate of hyperglycemia in the BBT was still lower compared with the CM group, but there was a higher risk of hypoglycemia in this group as well. Adjusted for other covariates, the major contributing factor driving the difference in hyperglycemia was steroid use and the number of outpatient antihyperglycemic medications, followed by differences in A1C and admission blood glucose.

Because of the retrospective nature of the study, patients selected for CM tended to be on fewer outpatient antihyperglycemic medications and had lower A1C and admission blood glucose values. This factor should be taken into consideration by providers to help guide their clinical decision-making as to which subset of patients should be initiated on BBT versus CM when admitted to the hospital.

All patients with diabetes or hyperglycemia should be monitored closely in the inpatient setting because hospitalized patients usually require alterations to outpatient diabetes management to maintain glycemia within target range (9). Capillary blood glucose monitoring is recommended in all patients with a known history of diabetes and in those without diabetes who have a blood glucose level >140 mg/dL. For hospitalized patients with diabetes who are eating on a regular schedule, POC blood glucose should be checked before each meal and at bedtime, if clinically indicated. For those who are not eating, POC blood glucose should be checked at least every 4–6 hours, with additional checks warranted for individuals at higher risk of hypoglycemia (10). This protocol maintains consistency, which is important because it is well established that sampling sources can affect glucose concentration measurements, with increasing concentrations in venous, capillary, and arterial samples, respectively (11).

Our study analyzed 994 blood glucose values and 2,118 blood glucose values in patients on BBT and CM, respectively, within the first 24 hours of admission from the onset of insulin administration. When extended out over 72 hours, 2,438 blood glucose values were obtained in patients on BBT and 5,490 blood glucose values were obtained in patients on CM. The percentages of euglycemia, hyperglycemia, and hypoglycemia among the blood glucose values obtained were reported in the BBT and CM groups.

Consistent with the ADA Standards of Medical Care in Diabetes (8), the policy at our institution is that oral and injectable noninsulin antihyperglycemic medications are discontinued and insulin therapy is initiated for all patients with diabetes who are admitted to the hospital. Insulin is the preferred agent for glycemic management in the inpatient setting because its pharmacodynamics allow it to be easily titrated based on glucose levels, food intake, and clinical condition, and it has minimal side effects and drug-drug interactions. Adults with diabetes managed with lifestyle modifications or noninsulin antihyperglycemic medications at home can initially be managed in the inpatient setting with CM or BBT to maintain blood glucose targets (12,13). For most noncritically ill patients in the hospital setting, a glucose target of 140–180 mg/dL is recommended, and a lower target of 100–140 mg/dL may be reasonable for those who are able to achieve and maintain this range without hypoglycemia (10). For patients being initiated on BBT, insulin can be initiated safely at a total daily dose of 0.2–0.6 units/kg body weight, divided into basal and bolus therapy, with careful consideration of age, kidney function, insulin resistance, admission blood glucose level, planned surgery, and corticosteroid use (14).

Several studies have demonstrated that sliding-scale insulin (SSI) is not acceptable as the sole regimen in patients with diabetes, yet it remains the most common default regimen in most institutions (15–18). Guidelines from the ADA for hospitalized patients with hyperglycemia requiring insulin discourage the use of CM and support the use of a basal-plus-correctional-insulin regimen, with the addition of prandial insulin for patients with consistent oral intake (8). The American Association of Clinical Endocrinologists recommends that the insulin regimen for hospitalized patients include basal, bolus, and correction doses and advises that exclusive use of SSI should only be for patients whose glucose levels are in the target range most of the time and only occasionally exceed that range (10). The use of scheduled insulin with basal and prandial insulins has proven to be an effective and safe strategy for glycemic management in the inpatient setting (19). CM may be appropriate when used for a short duration to assess total daily insulin dose requirements in patients without diabetes who have elevated blood glucose levels, individuals being initiated on corticosteroid therapy, and individuals on enteral or total parenteral nutrition, but it should not be used routinely, and hospitalized adults started on CM with persistent hyperglycemia (defined as two or more POC blood glucose values ≥180 mg/dL in a 24-hour period) should be switched to BBT (12).

Multiple studies have commented on the role that staff play in determining the effectiveness of an insulin regimen. In a prospective RCT by Said et al. (20) that compared SSI, 70/30 premixed insulin, and BBT, the authors partially attributed poor glycemic management in the BBT group to nursing errors that resulted from a lack of education and poorly trained nursing staff. The study noted that the multiple types of insulin caused confusion and resulted in nurse self-decision of giving only one type of insulin, which ultimately led to large fluctuations in blood glucose in the following days. A prospective study on implementation of a nurse-based basal-bolus regimen compared with a sliding-scale regimen in an orthopedic surgery unit demonstrated that average blood glucose and glucose variability were lower in the basal-bolus group, proving the feasibility and superiority of nurse-based basal-bolus versus SSI treatment in the context of intense training (21). Our study did not address issues of accurate communication between prescribing providers and nursing staff or correct administration of medications; however, our hospital has been an insulin-only institution for more than a decade, and managing BBT has been well engrained in our staff, who feel quite comfortable with its administration.

At our institution, providers are given clinical decision support within the EMR system when initiating insulin therapy but ultimately can choose whether to initiate BBT or CM. One of the more common reasons for the initiation of CM is to avoid hypoglycemia because, with CM, insulin is administered only if blood glucose levels are high. Although CM has been used to help determine correct doses for subsequent insulin administration, the use of CM in hospitalized patients has never been associated with improved clinical outcomes and has not been shown to prevent hyperglycemia; in fact, one study found that 84% of correctional insulin doses failed to correct hyperglycemia (22).

Our study results reinforce that lower rates of hyperglycemia without higher rates of hypoglycemia are found in patients on BBT compared with those on CM, although the clinical decision-making process itself is vulnerable to bias. For example, CM may be initiated over BBT in an effort to avoid hypoglycemia if a patient is presenting with poor caloric intake, is receiving nothing by mouth, has altered mental status, or is awaiting a surgical procedure. Therefore, the lower rates of hyperglycemia without higher rates of hypoglycemia in the BBT group compared with the CM group may be skewed by this clinical subjectivity. We recognize this unmeasured bias as inherent to the clinical decision-making process and as a limitation to the conclusions that can be drawn from our study.

Multiple previous studies have compared the use of BBT to CM, as well as comparing other inpatient insulin regimens. In the RABBIT 2 (Randomized Study of Basal-Bolus Insulin Therapy in the Inpatient Management of Patients With Type 2 Diabetes) trial (23), which was a prospective, multicenter, RCT conducted on medical inpatients in the United States, BBT was compared with SSI in hospitalized patients with type 2 diabetes. This study demonstrated that 66% of patients receiving BBT achieved glycemic targets compared with only 38% in the SSI group. There was no statistical difference in hypoglycemia or length of stay between the groups, indicating that BBT results in a significant improvement in glycemic management compared with the sole use of SSI. This was consistent with the RABBIT 2 Surgery (Randomized Study of Basal-Bolus Insulin Therapy in the Inpatient Management of Patients With Type 2 Diabetes Undergoing General Surgery) trial (24), which was a similarly conducted RCT involving surgical inpatients in the United States that demonstrated that 55% of patients receiving BBT achieved glycemic targets compared with only 31% in the SSI group. Hypoglycemia was reported in 23.1% of patients in the BBT group and 4.7% in the SSI group, with no significant differences in the frequency of severe hypoglycemia between the groups, leading the authors to conclude that BBT was preferred over SSI in the management of noncritically ill hospitalized patients with type 2 diabetes (24).

Comparatively, our study found there was no significant difference in rates of hypoglycemia between the BBT and CM groups within the first 24 hours of admission from the onset of insulin administration, with 1.9% of blood glucose values hypoglycemic in the BBT group compared with 1.4% in the CM group (P = 0.301). The proportion of patients who had hypoglycemic events was 4.0% in the BBT group and 4.3% in the CM group within the first 24 hours of admission from the onset of insulin administration. However, when extended to 72 hours, there was a significant difference in rates of hypoglycemia between the two groups, with 1.9% of blood glucose values hypoglycemic in the BBT group compared with 0.84% in the CM group (P <0.001). The proportion of patients with hypoglycemic events in this time frame was 8.8% in the BBT group and 5.3% in the CM group.

The Triple-B (Basal-Bolus-Booster) study by Perera et al. (25) was adapted from the RABBIT 2 trial with the aim to introduce a practical insulin protocol for hospitalized patients with hyperglycemia. Fifty-seven hospitalized patients with significant hyperglycemia were treated with a weight-based protocol consisting of basal insulin, bolus insulin, and booster (correctional) insulin, and their glycemic management was compared with 45 historical control subjects with similar age, A1C, and diabetes duration treated with only correctional insulin. This study demonstrated that the basal-bolus-booster group had significantly lower blood glucose levels and fewer episodes of hyperglycemia and that patients in the basal-bolus-booster group who experienced hypoglycemia were less likely to have a repeat episode.

The Basal-Plus study (26) further investigated inpatient use of basal insulin; this RCT compared regimens of basal-bolus insulin plus correctional insulin before meals (basal-bolus), basal insulin plus correctional insulin before meals (basal-plus), and SSI in patients with type 2 diabetes who were treated with lifestyle modification, oral antihyperglycemic medications, or low-dose insulin before admission. The study concluded that improvement in mean daily blood glucose after the first day of therapy was similar between basal-bolus and basal-plus groups and that both regimens resulted in lower mean daily blood glucose levels than SSI; in addition, it found that both basal regimens resulted in fewer occurrences of more than two consecutive blood glucose values >240 mg/dL or a mean daily blood glucose >240 mg/dL than did treatment with SSI. Hypoglycemia occurred in 16% of patients in the basal-bolus group, 13% in the basal-plus group, and 3% in the SSI group, with no difference among the groups in the frequency of severe hypoglycemia.

The Basal-Plus study (26) compared regimens, including basal insulin, to a regimen of SSI and assessed these factors over 24 hours of hospital admission, but it distinguished between the regimen of basal, bolus, and correctional insulin and the regimen of basal plus correctional insulin, whereas these regimens were grouped together and compared with the CM regimen in our study. The results of this study indicate that, even if our study were to have differentiated between those treated with basal, bolus, and correctional insulin versus those treated with basal and correctional insulin, the results likely would have been similar, as there were no significant differences in glycemic management between the two groups treated with basal insulin.

In contrast to these studies, which found improved glycemic management with basal insulin, a cross-sectional study by Roberts et al. (27), in which blood glucose levels were prospectively measured and treated with BBT and then compared with retrospective data from patients treated with SSI, demonstrated that mean blood glucose for patients receiving BBT was lower than baseline after the full first day of therapy and remained lower than baseline throughout the study, whereas there was no significant change in blood glucose for patients receiving SSI.

Our study did not measure the daily dose of insulin administered to each group; however, previous studies have compared insulin doses between BBT and CM. In a prospective study that compared SSI, 70/30 premixed insulin, and BBT, Said et al. (20) measured total insulin doses given and found that a significantly higher dose was administered to the subjects of the BBT group than the CM group. The RABBIT 2 trial also found a significantly higher mean daily insulin dose in the BBT group compared with the SSI group (23), and a higher total insulin dose was also found in the RABBIT 2 Surgery trial (24). It is inherent to a study comparing BBT and CM that higher total doses of insulin would be administered using a BBT regimen. A limitation of our study, however, is that we did not measure these total doses and therefore did not use multivariable analysis to account for possible differences in the amounts of insulin each group received.

Evidence from the majority of previous studies indicates that a regimen of basal, bolus, and correctional insulin achieves better glycemic management than correctional insulin alone. Although we analyzed glycemic assessments for 72 hours of admission from the onset of insulin administration, our study focused specifically on the first 24 hours of admission, which is a time period that is completely discounted in some studies and averaged with the rest of the admission values in others. This difference in our approach indicates that glycemic management during the first 24 hours of therapy may differ in key ways from the remainder of hospitalization stays.

Strengths and Limitations

At our institution, there is no set protocol for adjusting insulin doses; rather, dose adjustments are made by clinical decision-making of the primary care team, in some cases with the assistance of the endocrinology consultation service. Our study therefore more accurately represents the actual environment of a hospital than a scenario in which a study protocol was put in place for dose adjustments.

Our study also measured the percentages of euglycemia, hyperglycemia, and hypoglycemia rather than the mean blood glucose values measured by other studies. This difference allowed for a deeper analysis of the data without influence by outliers or skewed distribution; while a result of our approach is a limit in assessment of the degree of hyperglycemia, which would lead to grouping those with blood glucose levels only slightly above goal with those with severely elevated blood glucose values, it prevents erasure of large fluctuations in blood glucose values through averaging.

Furthermore, our study specifically assessed glycemic management in patients who were not treated with insulin at home and therefore could not be generalized to patients using insulin at home, whereas other studies included assessment of patients on home insulin.

Finally, our study was an observational retrospective study, which did not influence providers’ decisions on insulin dosing; whereas many previous studies were RCTs with specific protocols for adjusting insulin, insulin dosing in our study was done solely at the discretion of the provider and therefore may reflect a more accurate assessment of the actual practice of inpatient glycemic management.

Conclusion

BBT replicates normal pancreatic physiology, which involves basal secretion to regulate hepatic gluconeogenesis and bolus secretion to promote uptake into tissues. The initiation of BBT for glycemic management of insulin-naive patients in the inpatient setting, when adjusted for covariates, is associated with a lower degree of hyperglycemia without a higher rate of hypoglycemia compared with CM. Future studies could further examine the level of presenting hyperglycemia and other covariates that should prompt providers to choose BBT versus CM.