American Journal of Therapeutics 23, e805–e809 (2016)
Effect of Full Correction Versus Partial Correction of Elevated Blood Glucose in the Emergency Department on Hospital Length of Stay Michaela Johnson-Clague, PharmD, Jessica DiLeo, PharmD, Michael D. Katz, PharmD, and Asad E. Patanwala, PharmD*
There is limited information to guide the extent to which asymptomatic hyperglycemia needs to be corrected in patients presenting to the emergency department (ED) with unrelated complaints. The objective of this study was to compare full correction (FC) versus partial correction (PC) of elevated blood glucose in the ED on hospital length of stay. This was a retrospective cohort study conducted in an academic ED in the United States. Adult diabetic patients with hyperglycemia (blood glucose .200 mg/dL) in the ED who were treated with subcutaneous insulin were included. Patients were categorized based on the level of blood glucose control achieved within the first 24 hours from triage: (1) FC group for whom blood glucose ,200 mg/dL was achieved or (2) PC group for whom blood glucose remained $200 mg/dL. The primary outcome measure was a comparison of hospital length of stay between groups. A total of 161 patients were included in this study (FC 5 81, PC 5 80). There was no significant difference between hospital length of stay in the FC [3 days (interquartile range, 1–5 days)] and PC [3 days (interquartile range, 2–6 days)] groups (P 5 0.159). In the multivariate analysis, after adjusting for potential confounders, there was no significant association between level of correction and hospital length of stay (log-transformed) (coefficient 0.238; 95% confidence interval, 20.062 to 0.537; P 5 0.119; R2 5 13%). The extent of glucose correction was not associated with a decrease in hospital length of stay in diabetic patients with hyperglycemia in the ED. Keywords: blood glucose, hyperglycemia, emergency service, hospital, diabetes mellitus
INTRODUCTION Patients with acute illness who present to the emergency department (ED) due to a wide variety of chief complaints often have concurrent hyperglycemia that is unrelated to their primary diagnosis. This may be due to changes in glucose uptake and utilization, increased glucose production, decreased glycogenesis, glucose intolerance, and insulin resistance that occur
Department of Pharmacy Practice & Science, College of Pharmacy, University of Arizona, Tucson, AZ. The authors have no conflicts of interest to declare. *Address for correspondence: Associate Professor, 1295 N. Martin, PO Box 210202, Tucson, AZ 85721, USA. E-mail: patanwala@ pharmacy.arizona.edu
during acute physiologic stress.1,2 Inpatient hyperglycemia due to any cause is associated with poor outcomes.3 Thus, it is recommended that during hospitalization blood glucose should be controlled.3 However, there is limited guidance regarding the early management of hyperglycemia in the ED. In most scenarios in the ED, the management of hyperglycemia unrelated to the patients’ chief complaint is often considered to be secondary and less urgent than other diagnostic and therapeutic interventions. In addition, aggressive correction also poses a potential risk of hypoglycemia in patients who may already be unstable. It may also warrant need for additional monitoring, thereby potentially increasing cost and ED length of stay. Thus, treatment of hyperglycemia is commonly deferred or treated conservatively until patients are admitted to the hospital.
1075–2765 Copyright © 2014 Wolters Kluwer Health, Inc. All rights reserved.
www.americantherapeutics.com
Copyright © 2014 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
e806
At one institution, use of a hyperglycemia treatment protocol for diabetic patients in the ED resulted in a decrease in overall hospital length of stay.4 Patients with initial blood glucose of greater than 200 mg/dL upon ED triage were given subcutaneous insulin aspart every 2 hours to reduce blood glucose to less than 200 mg/dL. This study was hypothesis generating and suggested that blood glucose should be controlled early during hospitalization. The early management of hyperglycemia in the ED requires additional resources and the use of insulin, which is considered to be a high-risk medication.5 However, as institutions consider the adoption of such protocols, there is a need for additional studies to determine if this level of correction is necessary. The objective of this study was to compare the effect of full correction (FC) versus partial correction (PC) of elevated blood glucose in the ED on hospital length of stay. Given the findings from previous studies, we hypothesized that patients with diabetes who achieved FC (blood glucose less than 200 mg/dL) within 24 hours of ED triage would have a decreased overall hospital length of stay compared with patients with partial control.
MATERIALS AND METHODS Study design and setting This was a retrospective cohort study conducted in an academic, tertiary medical center ED in the United States between June 1, 2011 and June 20, 2012. The Institutional Review Board of the University and the Site Review Authority of the medical center approved the study. The ED did not have any protocol in place for blood glucose management, thus treatment was based on provider preference. This typically includes the use of subcutaneous rapid acting insulin (regular or lispro), with dosing and interval based on provider preference. Home medications, such as oral therapies, are held while in the ED. Selection of participants Consecutive patients who had diabetes and received subcutaneous rapid acting insulin (regular or lispro) in the ED for blood glucose correction were screened for inclusion. The intent was to include diabetic patients with high initial blood glucose value ($200 mg/dL), for whom the provider determined the need for treatment in the ED. Patients were excluded if they were younger than 18 years, discharged home directly from the ED, had diabetic ketoacidosis or hyperglycemic hyperosmolar state, treated with insulin for hyperkalemia, trauma patients or nondiabetic patients, or patients with American Journal of Therapeutics (2016) 23(3)
Johnson-Clague et al
initial blood glucose levels ,200 mg/dL. All blood glucose correction was performed using subcutaneous insulin, and none of the patients received insulin infusions. We only included patients with diabetes rather than patients with incidental hyperglycemia to be consistent with a previous study in this setting.4 Patients were categorized into 2 groups based on whether a blood glucose value of ,200 mg/dL was achieved within 24 hours of triage in the ED: (1) FC group for whom blood glucose ,200 mg/dL was achieved and (2) PC group for whom blood glucose remained $200 mg/dL. This cut off value of 200 mg/dL was chosen based on a previous study in this setting.4 The occurrence of hypoglycemia was recorded. This was defined as blood glucose ,60 mg/dL. Data collection A standardized data collection form was used to obtain data from medical records. Variables collected included patient demographics, number of diagnoses, insulin administered in the ED (dose and type), and hemoglobin A1C. The primary outcome measure was hospital length of stay, which was also collected. The extent of comorbidities was quantified by calculating the Charlson Comorbidity Index (CCI).6,7 An increasing index indicates a higher level of comorbid conditions. Severity of illness upon presentation to the ED was estimated by calculating the Rapid Emergency Medicine Score (REMS).8 The REMS is a validated measure that is calculated from variables such as age, vital signs, oxygenation, and neurological assessment. An increasing score also indicates a higher severity of illness. All finger stick blood glucose measurements during the initial 24 hours from patient presentation to the ED was obtained. Data analysis Continuous variables, including hospital length of stay, were compared between the 2 groups using the Wilcoxon rank sum test and reported as medians with interquartile ranges (IQRs). Categorical variables were compared using the Fisher’s exact test and reported as percentages. A multivariate linear regression analysis was performed to determine the effect of blood glucose control on hospital length of stay after adjusting for potential confounders. Potential confounders were determined based on previous literature and clinical experience of the investigators. These were age, sex, race, body mass index, hemoglobin A1C, CCI, number of diagnoses, REMS, and diabetes type (1 or 2). Age was not included in the model because it is already part of the REMS. CCI and number of diagnoses are both indicators of comorbidity. Thus, the variable with the strongest association with hospital length of stay was entered into the model. These variables and the REMS were entered into the model as www.americantherapeutics.com
Copyright © 2014 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Blood Glucose in the Emergency Department
e807
ordinal variables. In the multivariate analysis, race was dichotomized as white or non-white. Hospital length of stay was log-transformed to meet model assumptions of linearity and homoscedasticity. The model was checked for interactions and residuals were generated and evaluated for normality. Four sensitivity analyses were conducted. First, patients who died in the hospital were excluded because early deaths could influence length of stay measures. Second, potential outliers (Cook’s distance .4/N) were excluded to determine if it changed the results. Third, the groups were categorized based on achieving blood glucose ,200 mg/dL within 12 hours rather than 24 hours. Fourth, because a large proportion of patients had missing hemoglobin A1C, this variable was included only as part of a sensitivity analysis. A power analysis was conducted assuming a mean length of stay of 3.8 and 5.3 days in the 2 groups, SD of 3 days each, alpha 0.05, and power of 80%. Using these parameters from the study by Munoz et al,4 a sample size of 126 was required (63 in each group). An a priori alpha of 0.05 was used for all analyses. All statistical evaluations were performed using Stata 13 (Stata Corporation, College Station, TX).
RESULTS Characteristics of study subjects A total of 475 medical records were evaluated. Of these, 99 had diabetic ketoacidosis or hyperglycemia hyperosmolar state requiring treatment with an insulin infusion, 89 were treated with insulin for hyperkalemia, 76 had an initial blood glucose less than 200 mg/dL, 37 were patients with trauma or nondiabetic patients, and 13 patients were not admitted to the hospital. Thus, 161 diabetic patients were included in this final study cohort (FC group 5 81, PC group 5 80). Overall, the median age was 55 years (IQR, 42–65 years) and most were male patients (56%, n 5 90). The most common race was Hispanic (49%, n 5 79), followed by white (35%, n 5 57), African American (3%, n 5 5), and other or missing (12%, n 5 20). Also, 61% (n 5 98) of patients were on chronic insulin therapy for the management of diabetes. Most patients had type 2 diabetes (91%, n 5 147) and a few had type 1 diabetes (9%, n 5 14). Baseline demographic and clinical comparisons between groups are reported in Table 1.
Table 1. Baseline patient characteristics by group. Patient characteristic Age (yrs) Sex (male %) Race/ethnicity (%) Hispanic White African American Other Diabetes type (%) Type 1 Type 2 Home medications (%) Insulin Metformin Sulfonylurea Initial blood glucose (mg/dL) Lowest blood glucose within 24 hours (mg/dL) Long-acting insulin (glargine) used in the ED (%) REMS CCI Number of diagnoses Body mass index (kg/m2) Hemoglobin A1C www.americantherapeutics.com
FC median (IQR) or N (%)
PC median (IQR) or N (%)
57 (48–66) 48 (59)
49 (40–62) 42 (53)
39 27 5 10
(48) (33) (6) (13)
40 (50) 30 (38) 0 (0) 10 (12)
7 (9) 74 (91)
7 (9) 73 (91)
P 0.008 0.429 0.171
1.000
47 31 17 267 148 17 5 3 14 30 9.0
(58) (38) (21) (232–379) (117–172) (21) (3–7) (1–5) (10–18) (25–37) (7.6–10.7)
51 27 11 402 250 16 4 2 13 33 10.7
(64) (34) (14) (318–479) (217–279) (20) (2–6) (1–4) (9–17) (26–39) (9.1–12.7)
0.519 0.623 0.299 ,0.001 ,0.001 1.000 0.015 0.104 0.491 0.158 ,0.001
American Journal of Therapeutics (2016) 23(3)
Copyright © 2014 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
e808
Johnson-Clague et al
DISCUSSION
Main results There was no significant difference between hospital length of stay in the FC [3 days (IQR, 1–5 days)] and PC [3 days (IQR, 2–6 days)] groups (P 5 0.159). In the multivariate analysis, after adjusting for potential confounders, there was no significant association between glucose control level and hospital length of stay (log-transformed, Table 2) (coefficient 5 0.238; 95% CI, 20.062–0.537; P 5 0.119; R2 5 13%). Four sensitivity analyses were conducted to assess if hospital deaths, outliers, categorization based on earlier control (within 12 hours), or inclusion of hemoglobin A1C changed the results. There were 2 patients who died while hospitalized. Both patients were in the FC group. Excluding these patients or potential outliers did not change the results. There were 38 patients in the FC group who achieved blood glucose ,200 mg/dL within 12 hours of ED triage. Categorization based on earlier control also did not change the results. A large proportion of the sample was missing hemoglobin A1C values (n 5 67). Thus, this variable was not included in the final model. However, the model was also run after including hemoglobin A1C, as a complete case analysis. Inclusion of this variable did not change the results. There were 2 patients categorized in the FC group who developed hypoglycemia (blood glucose ,60 mg/dL). Both had hypoglycemia (45 and 59 mg/dL) after transfer from the ED but within 24 hours. No clinical consequences of hypoglycemia were identified. Table 2. Multivariate analysis. Patient characteristic Glucose control FC PC Sex Male Female Race Non-white White Diabetes type Type 2 Type 1 Body mass index (kg/m2) REMS Number of diagnoses
Coefficient
95% CI
P
[Reference] 0.238 20.062 to 0.537
0.119
[Reference] 20.153 20.455 to 0.149
0.317
[Reference] 0.142 20.178 to 0.461
0.383
[Reference] 0.047 20.505 to 0.599 20.002 20.016 to 0.013
0.868 0.821
20.022 0.045
20.084 to 0.040 0.020 to 0.069
American Journal of Therapeutics (2016) 23(3)
0.480 ,0.001
The key finding from the investigation was that level of glucose control was not associated with overall hospital length of stay in diabetic patients. These results are different than the study by Munoz et al,4 in which the impact of a subcutaneous insulin protocol in the ED showed a decreased length of hospital stay. In the study by Munoz et al, 54 diabetic patients were given insulin aspart every 2 hours in the ED until blood glucose of ,200 mg/dL was achieved. These patients were then compared with a historical control group, with regard to hospital length of stay. However, this was a relatively small study, and it is unclear if patient matching with the historical control was optimal. The investigators did report a categorical severity of illness measure that is derived from administrative data, but level of comorbidity was not measured or reported. In our study, we used the REMS, which has been validated in the ED setting and uses clinical data for calculation. In addition, comorbidity was measured using the CCI and number of diagnoses. We also performed a multivariate analysis to adjust for important confounders, which was not conducted in the previous study.4 Our study adds to the study by Munoz et al4 because the results suggest that PC of blood glucose may be sufficient without a deleterious effect on patient outcomes. Bernard et al9 evaluated the impact of an inpatient hyperglycemia protocol that was initiated in the ED. This was a randomized trial involving 176 patients with diabetes. The treatment group received aspart insulin for hyperglycemia in the ED, which was titrated to achieve blood glucose ,200 mg/dL. Patients then received daily detemir insulin during hospitalization. The control group received usual care. The authors found no difference in hospital length of stay, which is consistent with our findings. However, the treatment group did have better glucose control during hospitalization. Based on this the authors advocated for standardization of insulin protocols during hospitalization. However, it is still unclear if early glucose control in the ED provided any additional benefit. As this was a retrospective analysis, data collection was dependent on accurate documentation in the medical record. It is possible that there were unmeasured confounders that could have influenced hospital length of stay. This could include any events after the first 24 hours during hospitalization. However, previous studies have indicated the benefit of early glucose control protocols as an independent treatment modality with a reduction in length of stay.4 Nonetheless, we adjusted for important confounders such as level of comorbidity and severity of illness. The groups were categorized www.americantherapeutics.com
Copyright © 2014 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Blood Glucose in the Emergency Department
based on glucose control at the 24-hour time point, which was somewhat arbitrary. But as part of a sensitivity analysis, the groups were also categorized at 12 hours and it did not change the results. Note that 12 hours was the median ED length of stay. This approach was preferred to categorizing patients based on glucose control within the ED because patients have different ED lengths of stay. Thus, there is better physiologic rationale to use consistent durations for categorization of all patients. We did not identify or include patients who had elevated blood glucose on ED triage and did not receive rapid acting insulin. These patients could have potentially been included in the PC group. However, the intent of the study was to include patients for whom an attempt had been made to correct hyperglycemia. Exclusion of untreated patients reduces the likelihood of selection bias in this observational study because the lack of treatment may be indicative of less severely ill patients or other unmeasured confounders. The FC group was slightly older than the PC group. But this difference is not considered to be clinically meaningful. Age is incorporated in the REMS score, which was part of our multivariate analysis. Initial blood glucose value was also significantly higher in the PC group. This is likely because higher blood glucose values are more difficult to correct. There were significant differences in REMS score and hemoglobin A1C. However, we adjusted for these variables in the multivariate and sensitivity analyses. In the final multivariate analyses, greater number of diagnoses were associated with greater length of stay. This highlights the importance of incorporation of this variable in the model. In conclusion, in diabetic patients who receive subcutaneous insulin for hyperglycemia in the ED, achieving a blood glucose value of ,200 mg/dL within 24 hours was not associated with a decrease in hospital length of stay. Thus, given limited resources in the ED setting, it may be appropriate to use a conservative approach for asymptomatic hyperglycemia until the patient is hospitalized, while more emergent chief
www.americantherapeutics.com
e809
complaints are addressed in the ED. Prospective studies are needed to confirm these findings.
ACKNOWLEDGMENTS The authors thank Jennifer Prze for her assistance with data collection.
REFERENCES 1. Mizock BA. Alterations in carbohydrate metabolism during stress: a review of the literature. Am J Med. 1995;98: 75–84. 2. Dungan KM, Braithwaite SS, Preiser JC. Stress hyperglycaemia. Lancet. 2009;373:1798–1807. 3. Moghissi ES, Korytkowski MT, DiNardo M, et al. American Association of Clinical Endocrinologists and American Diabetes Association consensus statement on inpatient glycemic control. Endocr Pract. 2009;15:353–369. 4. Munoz C, Villanueva G, Fogg L, et al. Impact of a subcutaneous insulin protocol in the emergency department: Rush Emergency Department Hyperglycemia Intervention (REDHI). J Emerg Med. 2011;40:493–498. 5. ISMP’s list of high alert medications. Available at: http://www.ismp.org/tools/institutionalhighAlert.asp. Accessed February 1, 2014. 6. Charlson ME, Pompei P, Ales KL, et al. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40:373–383. 7. Olsson T, Terent A, Lind L. Charlson Comorbidity Index can add prognostic information to Rapid Emergency Medicine Score as a predictor of long-term mortality. Eur J Emerg Med. 2005;12:220–224. 8. Olsson T, Terent A, Lind L. Rapid Emergency Medicine score: a new prognostic tool for in-hospital mortality in nonsurgical emergency department patients. J Intern Med. 2004;255:579–587. 9. Bernard JB, Munoz C, Harper J, et al. Treatment of inpatient hyperglycemia beginning in the emergency department: a randomized trial using insulins aspart and detemir compared with usual care. J Hosp Med. 2011;6:279–284.
American Journal of Therapeutics (2016) 23(3)
Copyright © 2014 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Effect of Full Correction Versus Partial Correction of Elevated Blood Glucose in the Emergency Department on Hospital Length of Stay Michaela Johnson-Clague, PharmD, Jessica DiLeo, PharmD, Michael D. Katz, PharmD, and Asad E. Patanwala, PharmD*
There is limited information to guide the extent to which asymptomatic hyperglycemia needs to be corrected in patients presenting to the emergency department (ED) with unrelated complaints. The objective of this study was to compare full correction (FC) versus partial correction (PC) of elevated blood glucose in the ED on hospital length of stay. This was a retrospective cohort study conducted in an academic ED in the United States. Adult diabetic patients with hyperglycemia (blood glucose .200 mg/dL) in the ED who were treated with subcutaneous insulin were included. Patients were categorized based on the level of blood glucose control achieved within the first 24 hours from triage: (1) FC group for whom blood glucose ,200 mg/dL was achieved or (2) PC group for whom blood glucose remained $200 mg/dL. The primary outcome measure was a comparison of hospital length of stay between groups. A total of 161 patients were included in this study (FC 5 81, PC 5 80). There was no significant difference between hospital length of stay in the FC [3 days (interquartile range, 1–5 days)] and PC [3 days (interquartile range, 2–6 days)] groups (P 5 0.159). In the multivariate analysis, after adjusting for potential confounders, there was no significant association between level of correction and hospital length of stay (log-transformed) (coefficient 0.238; 95% confidence interval, 20.062 to 0.537; P 5 0.119; R2 5 13%). The extent of glucose correction was not associated with a decrease in hospital length of stay in diabetic patients with hyperglycemia in the ED. Keywords: blood glucose, hyperglycemia, emergency service, hospital, diabetes mellitus
INTRODUCTION Patients with acute illness who present to the emergency department (ED) due to a wide variety of chief complaints often have concurrent hyperglycemia that is unrelated to their primary diagnosis. This may be due to changes in glucose uptake and utilization, increased glucose production, decreased glycogenesis, glucose intolerance, and insulin resistance that occur
Department of Pharmacy Practice & Science, College of Pharmacy, University of Arizona, Tucson, AZ. The authors have no conflicts of interest to declare. *Address for correspondence: Associate Professor, 1295 N. Martin, PO Box 210202, Tucson, AZ 85721, USA. E-mail: patanwala@ pharmacy.arizona.edu
during acute physiologic stress.1,2 Inpatient hyperglycemia due to any cause is associated with poor outcomes.3 Thus, it is recommended that during hospitalization blood glucose should be controlled.3 However, there is limited guidance regarding the early management of hyperglycemia in the ED. In most scenarios in the ED, the management of hyperglycemia unrelated to the patients’ chief complaint is often considered to be secondary and less urgent than other diagnostic and therapeutic interventions. In addition, aggressive correction also poses a potential risk of hypoglycemia in patients who may already be unstable. It may also warrant need for additional monitoring, thereby potentially increasing cost and ED length of stay. Thus, treatment of hyperglycemia is commonly deferred or treated conservatively until patients are admitted to the hospital.
1075–2765 Copyright © 2014 Wolters Kluwer Health, Inc. All rights reserved.
www.americantherapeutics.com
Copyright © 2014 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
e806
At one institution, use of a hyperglycemia treatment protocol for diabetic patients in the ED resulted in a decrease in overall hospital length of stay.4 Patients with initial blood glucose of greater than 200 mg/dL upon ED triage were given subcutaneous insulin aspart every 2 hours to reduce blood glucose to less than 200 mg/dL. This study was hypothesis generating and suggested that blood glucose should be controlled early during hospitalization. The early management of hyperglycemia in the ED requires additional resources and the use of insulin, which is considered to be a high-risk medication.5 However, as institutions consider the adoption of such protocols, there is a need for additional studies to determine if this level of correction is necessary. The objective of this study was to compare the effect of full correction (FC) versus partial correction (PC) of elevated blood glucose in the ED on hospital length of stay. Given the findings from previous studies, we hypothesized that patients with diabetes who achieved FC (blood glucose less than 200 mg/dL) within 24 hours of ED triage would have a decreased overall hospital length of stay compared with patients with partial control.
MATERIALS AND METHODS Study design and setting This was a retrospective cohort study conducted in an academic, tertiary medical center ED in the United States between June 1, 2011 and June 20, 2012. The Institutional Review Board of the University and the Site Review Authority of the medical center approved the study. The ED did not have any protocol in place for blood glucose management, thus treatment was based on provider preference. This typically includes the use of subcutaneous rapid acting insulin (regular or lispro), with dosing and interval based on provider preference. Home medications, such as oral therapies, are held while in the ED. Selection of participants Consecutive patients who had diabetes and received subcutaneous rapid acting insulin (regular or lispro) in the ED for blood glucose correction were screened for inclusion. The intent was to include diabetic patients with high initial blood glucose value ($200 mg/dL), for whom the provider determined the need for treatment in the ED. Patients were excluded if they were younger than 18 years, discharged home directly from the ED, had diabetic ketoacidosis or hyperglycemic hyperosmolar state, treated with insulin for hyperkalemia, trauma patients or nondiabetic patients, or patients with American Journal of Therapeutics (2016) 23(3)
Johnson-Clague et al
initial blood glucose levels ,200 mg/dL. All blood glucose correction was performed using subcutaneous insulin, and none of the patients received insulin infusions. We only included patients with diabetes rather than patients with incidental hyperglycemia to be consistent with a previous study in this setting.4 Patients were categorized into 2 groups based on whether a blood glucose value of ,200 mg/dL was achieved within 24 hours of triage in the ED: (1) FC group for whom blood glucose ,200 mg/dL was achieved and (2) PC group for whom blood glucose remained $200 mg/dL. This cut off value of 200 mg/dL was chosen based on a previous study in this setting.4 The occurrence of hypoglycemia was recorded. This was defined as blood glucose ,60 mg/dL. Data collection A standardized data collection form was used to obtain data from medical records. Variables collected included patient demographics, number of diagnoses, insulin administered in the ED (dose and type), and hemoglobin A1C. The primary outcome measure was hospital length of stay, which was also collected. The extent of comorbidities was quantified by calculating the Charlson Comorbidity Index (CCI).6,7 An increasing index indicates a higher level of comorbid conditions. Severity of illness upon presentation to the ED was estimated by calculating the Rapid Emergency Medicine Score (REMS).8 The REMS is a validated measure that is calculated from variables such as age, vital signs, oxygenation, and neurological assessment. An increasing score also indicates a higher severity of illness. All finger stick blood glucose measurements during the initial 24 hours from patient presentation to the ED was obtained. Data analysis Continuous variables, including hospital length of stay, were compared between the 2 groups using the Wilcoxon rank sum test and reported as medians with interquartile ranges (IQRs). Categorical variables were compared using the Fisher’s exact test and reported as percentages. A multivariate linear regression analysis was performed to determine the effect of blood glucose control on hospital length of stay after adjusting for potential confounders. Potential confounders were determined based on previous literature and clinical experience of the investigators. These were age, sex, race, body mass index, hemoglobin A1C, CCI, number of diagnoses, REMS, and diabetes type (1 or 2). Age was not included in the model because it is already part of the REMS. CCI and number of diagnoses are both indicators of comorbidity. Thus, the variable with the strongest association with hospital length of stay was entered into the model. These variables and the REMS were entered into the model as www.americantherapeutics.com
Copyright © 2014 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Blood Glucose in the Emergency Department
e807
ordinal variables. In the multivariate analysis, race was dichotomized as white or non-white. Hospital length of stay was log-transformed to meet model assumptions of linearity and homoscedasticity. The model was checked for interactions and residuals were generated and evaluated for normality. Four sensitivity analyses were conducted. First, patients who died in the hospital were excluded because early deaths could influence length of stay measures. Second, potential outliers (Cook’s distance .4/N) were excluded to determine if it changed the results. Third, the groups were categorized based on achieving blood glucose ,200 mg/dL within 12 hours rather than 24 hours. Fourth, because a large proportion of patients had missing hemoglobin A1C, this variable was included only as part of a sensitivity analysis. A power analysis was conducted assuming a mean length of stay of 3.8 and 5.3 days in the 2 groups, SD of 3 days each, alpha 0.05, and power of 80%. Using these parameters from the study by Munoz et al,4 a sample size of 126 was required (63 in each group). An a priori alpha of 0.05 was used for all analyses. All statistical evaluations were performed using Stata 13 (Stata Corporation, College Station, TX).
RESULTS Characteristics of study subjects A total of 475 medical records were evaluated. Of these, 99 had diabetic ketoacidosis or hyperglycemia hyperosmolar state requiring treatment with an insulin infusion, 89 were treated with insulin for hyperkalemia, 76 had an initial blood glucose less than 200 mg/dL, 37 were patients with trauma or nondiabetic patients, and 13 patients were not admitted to the hospital. Thus, 161 diabetic patients were included in this final study cohort (FC group 5 81, PC group 5 80). Overall, the median age was 55 years (IQR, 42–65 years) and most were male patients (56%, n 5 90). The most common race was Hispanic (49%, n 5 79), followed by white (35%, n 5 57), African American (3%, n 5 5), and other or missing (12%, n 5 20). Also, 61% (n 5 98) of patients were on chronic insulin therapy for the management of diabetes. Most patients had type 2 diabetes (91%, n 5 147) and a few had type 1 diabetes (9%, n 5 14). Baseline demographic and clinical comparisons between groups are reported in Table 1.
Table 1. Baseline patient characteristics by group. Patient characteristic Age (yrs) Sex (male %) Race/ethnicity (%) Hispanic White African American Other Diabetes type (%) Type 1 Type 2 Home medications (%) Insulin Metformin Sulfonylurea Initial blood glucose (mg/dL) Lowest blood glucose within 24 hours (mg/dL) Long-acting insulin (glargine) used in the ED (%) REMS CCI Number of diagnoses Body mass index (kg/m2) Hemoglobin A1C www.americantherapeutics.com
FC median (IQR) or N (%)
PC median (IQR) or N (%)
57 (48–66) 48 (59)
49 (40–62) 42 (53)
39 27 5 10
(48) (33) (6) (13)
40 (50) 30 (38) 0 (0) 10 (12)
7 (9) 74 (91)
7 (9) 73 (91)
P 0.008 0.429 0.171
1.000
47 31 17 267 148 17 5 3 14 30 9.0
(58) (38) (21) (232–379) (117–172) (21) (3–7) (1–5) (10–18) (25–37) (7.6–10.7)
51 27 11 402 250 16 4 2 13 33 10.7
(64) (34) (14) (318–479) (217–279) (20) (2–6) (1–4) (9–17) (26–39) (9.1–12.7)
0.519 0.623 0.299 ,0.001 ,0.001 1.000 0.015 0.104 0.491 0.158 ,0.001
American Journal of Therapeutics (2016) 23(3)
Copyright © 2014 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
e808
Johnson-Clague et al
DISCUSSION
Main results There was no significant difference between hospital length of stay in the FC [3 days (IQR, 1–5 days)] and PC [3 days (IQR, 2–6 days)] groups (P 5 0.159). In the multivariate analysis, after adjusting for potential confounders, there was no significant association between glucose control level and hospital length of stay (log-transformed, Table 2) (coefficient 5 0.238; 95% CI, 20.062–0.537; P 5 0.119; R2 5 13%). Four sensitivity analyses were conducted to assess if hospital deaths, outliers, categorization based on earlier control (within 12 hours), or inclusion of hemoglobin A1C changed the results. There were 2 patients who died while hospitalized. Both patients were in the FC group. Excluding these patients or potential outliers did not change the results. There were 38 patients in the FC group who achieved blood glucose ,200 mg/dL within 12 hours of ED triage. Categorization based on earlier control also did not change the results. A large proportion of the sample was missing hemoglobin A1C values (n 5 67). Thus, this variable was not included in the final model. However, the model was also run after including hemoglobin A1C, as a complete case analysis. Inclusion of this variable did not change the results. There were 2 patients categorized in the FC group who developed hypoglycemia (blood glucose ,60 mg/dL). Both had hypoglycemia (45 and 59 mg/dL) after transfer from the ED but within 24 hours. No clinical consequences of hypoglycemia were identified. Table 2. Multivariate analysis. Patient characteristic Glucose control FC PC Sex Male Female Race Non-white White Diabetes type Type 2 Type 1 Body mass index (kg/m2) REMS Number of diagnoses
Coefficient
95% CI
P
[Reference] 0.238 20.062 to 0.537
0.119
[Reference] 20.153 20.455 to 0.149
0.317
[Reference] 0.142 20.178 to 0.461
0.383
[Reference] 0.047 20.505 to 0.599 20.002 20.016 to 0.013
0.868 0.821
20.022 0.045
20.084 to 0.040 0.020 to 0.069
American Journal of Therapeutics (2016) 23(3)
0.480 ,0.001
The key finding from the investigation was that level of glucose control was not associated with overall hospital length of stay in diabetic patients. These results are different than the study by Munoz et al,4 in which the impact of a subcutaneous insulin protocol in the ED showed a decreased length of hospital stay. In the study by Munoz et al, 54 diabetic patients were given insulin aspart every 2 hours in the ED until blood glucose of ,200 mg/dL was achieved. These patients were then compared with a historical control group, with regard to hospital length of stay. However, this was a relatively small study, and it is unclear if patient matching with the historical control was optimal. The investigators did report a categorical severity of illness measure that is derived from administrative data, but level of comorbidity was not measured or reported. In our study, we used the REMS, which has been validated in the ED setting and uses clinical data for calculation. In addition, comorbidity was measured using the CCI and number of diagnoses. We also performed a multivariate analysis to adjust for important confounders, which was not conducted in the previous study.4 Our study adds to the study by Munoz et al4 because the results suggest that PC of blood glucose may be sufficient without a deleterious effect on patient outcomes. Bernard et al9 evaluated the impact of an inpatient hyperglycemia protocol that was initiated in the ED. This was a randomized trial involving 176 patients with diabetes. The treatment group received aspart insulin for hyperglycemia in the ED, which was titrated to achieve blood glucose ,200 mg/dL. Patients then received daily detemir insulin during hospitalization. The control group received usual care. The authors found no difference in hospital length of stay, which is consistent with our findings. However, the treatment group did have better glucose control during hospitalization. Based on this the authors advocated for standardization of insulin protocols during hospitalization. However, it is still unclear if early glucose control in the ED provided any additional benefit. As this was a retrospective analysis, data collection was dependent on accurate documentation in the medical record. It is possible that there were unmeasured confounders that could have influenced hospital length of stay. This could include any events after the first 24 hours during hospitalization. However, previous studies have indicated the benefit of early glucose control protocols as an independent treatment modality with a reduction in length of stay.4 Nonetheless, we adjusted for important confounders such as level of comorbidity and severity of illness. The groups were categorized www.americantherapeutics.com
Copyright © 2014 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Blood Glucose in the Emergency Department
based on glucose control at the 24-hour time point, which was somewhat arbitrary. But as part of a sensitivity analysis, the groups were also categorized at 12 hours and it did not change the results. Note that 12 hours was the median ED length of stay. This approach was preferred to categorizing patients based on glucose control within the ED because patients have different ED lengths of stay. Thus, there is better physiologic rationale to use consistent durations for categorization of all patients. We did not identify or include patients who had elevated blood glucose on ED triage and did not receive rapid acting insulin. These patients could have potentially been included in the PC group. However, the intent of the study was to include patients for whom an attempt had been made to correct hyperglycemia. Exclusion of untreated patients reduces the likelihood of selection bias in this observational study because the lack of treatment may be indicative of less severely ill patients or other unmeasured confounders. The FC group was slightly older than the PC group. But this difference is not considered to be clinically meaningful. Age is incorporated in the REMS score, which was part of our multivariate analysis. Initial blood glucose value was also significantly higher in the PC group. This is likely because higher blood glucose values are more difficult to correct. There were significant differences in REMS score and hemoglobin A1C. However, we adjusted for these variables in the multivariate and sensitivity analyses. In the final multivariate analyses, greater number of diagnoses were associated with greater length of stay. This highlights the importance of incorporation of this variable in the model. In conclusion, in diabetic patients who receive subcutaneous insulin for hyperglycemia in the ED, achieving a blood glucose value of ,200 mg/dL within 24 hours was not associated with a decrease in hospital length of stay. Thus, given limited resources in the ED setting, it may be appropriate to use a conservative approach for asymptomatic hyperglycemia until the patient is hospitalized, while more emergent chief
www.americantherapeutics.com
e809
complaints are addressed in the ED. Prospective studies are needed to confirm these findings.
ACKNOWLEDGMENTS The authors thank Jennifer Prze for her assistance with data collection.
REFERENCES 1. Mizock BA. Alterations in carbohydrate metabolism during stress: a review of the literature. Am J Med. 1995;98: 75–84. 2. Dungan KM, Braithwaite SS, Preiser JC. Stress hyperglycaemia. Lancet. 2009;373:1798–1807. 3. Moghissi ES, Korytkowski MT, DiNardo M, et al. American Association of Clinical Endocrinologists and American Diabetes Association consensus statement on inpatient glycemic control. Endocr Pract. 2009;15:353–369. 4. Munoz C, Villanueva G, Fogg L, et al. Impact of a subcutaneous insulin protocol in the emergency department: Rush Emergency Department Hyperglycemia Intervention (REDHI). J Emerg Med. 2011;40:493–498. 5. ISMP’s list of high alert medications. Available at: http://www.ismp.org/tools/institutionalhighAlert.asp. Accessed February 1, 2014. 6. Charlson ME, Pompei P, Ales KL, et al. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40:373–383. 7. Olsson T, Terent A, Lind L. Charlson Comorbidity Index can add prognostic information to Rapid Emergency Medicine Score as a predictor of long-term mortality. Eur J Emerg Med. 2005;12:220–224. 8. Olsson T, Terent A, Lind L. Rapid Emergency Medicine score: a new prognostic tool for in-hospital mortality in nonsurgical emergency department patients. J Intern Med. 2004;255:579–587. 9. Bernard JB, Munoz C, Harper J, et al. Treatment of inpatient hyperglycemia beginning in the emergency department: a randomized trial using insulins aspart and detemir compared with usual care. J Hosp Med. 2011;6:279–284.
American Journal of Therapeutics (2016) 23(3)
Copyright © 2014 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
![[Length of stay in the hospital emergency department].](/assets/img/hold.png)
