Abstract and Introduction

Abstract

Study Objectives: To evaluate the effectiveness and safety of maintaining a target blood glucose concentration of 91–130 mg/dl with a standardized, nurse-managed, intensive insulin infusion protocol outside a study setting, and to determine if a statistically significant favorable effect on morbidity and mortality was achieved.
Design: Retrospective, observational, chart review.
Setting: Medical and surgical intensive care units (ICUs) in a community teaching hospital.
Patients: One hundred forty-three adult patients who received insulin infusions managed at the discretion of the physician over a 1-year period before initiation of the protocol (control group), and 70 patients who received insulin infusions over a 6-month period with infusion dosages titrated by using the protocol (protocol group).
Measurements and Main Results: Episodes of hypoglycemia, time within target range, mean blood glucose concentration, frequency of measurement, length of ICU stay, duration of mechanical ventilation, and overall mortality were collected. Hypoglycemic episodes were not significantly different between the groups. Blood glucose concentrations were within target range in 34% of all measurements in the protocol group compared with 23% in the control group (p < 0.001, relative risk [RR] 1.48, 95% confidence interval [CI] 1.38–1.58). Once target range was reached on one measurement, 43% of concentrations remained in target range in the protocol group compared with 29% in the control group (p < 0.001, RR 1.47, 95% CI 1.38–1.56). Frequency of measurements was higher in the protocol group versus control group (p=0.01); however, clinical difference was minimal. Protocol group had lower overall mortality rate (27% [19/70] vs 32% [46/143], p = 0.45), reduced mean ICU length of stay (16.7 - 10.6 vs 18.4 ? 16.0 days, p = 0.37), and less mechanical ventilation time (16.5 ? 9.7 vs 17.0 ? 15.0 days, p = 0.79).
Conclusion: The nurse-managed insulin infusion protocol improved glycemic control with minimal hypoglycemic episodes compared with baseline practice. A trend toward decreased mortality, ICU length of stay, and days of mechanical ventilation was observed. When compared with other published protocols, our insulin protocol displays comparable effectiveness with the use of less-frequent blood glucose measurements.

Introduction

The management of blood glucose concentrations in the intensive care unit (ICU) has gained much attention since a report published in 2001 indicated that maintaining a blood glucose concentration of 80–110 mg/dl in the critically ill patient leads to an overall reduction in morbidity and mortality.^[1] As a result of this research, ICUs have been challenged to duplicate those same results, outside the research setting, in medical, surgical, and neurologic ICUs.

Individualizing insulin infusions is time consuming for both physicians and nursing staff and, at times, leads to delays in optimization of blood glucose concentrations. Therefore, to better achieve tight blood glucose control in the ICU, a protocol-driven approach appears to be the most practical. The above-mentioned published report^[1] described a protocol, performed by research nurses dedicated solely to this function, that was effective in achieving a blood glucose goal of 80–110 mg/dl.^[1] Effectiveness of this protocol was measured by evaluating one morning blood glucose concentration. In assessing the effectiveness of insulin protocols outside the research setting, it is unclear if a morning blood glucose measurement is an accurate representation on how well a protocol performs over the course of the day. In addition, the authors of the above report^[1] defined hypoglycemia as a blood glucose below 40 mg/dl, a concentration that triggers dextrose administration in most ICUs. Because glucose counter-regulatory systems are triggered at a plasma glucose of 65–70 mg/dl in individuals who are not diabetic, hypoglycemia is defined by the American Diabetes Association as a blood glucose of 70 mg/dl or less, with or without typical symptoms of hypoglycemia.^[2] Evaluating the safety of an insulin protocol may be better observed at this threshold.

Studies on the Portland continuous intravenous infusion protocol most notably reported an abundance of data on the success of achieving blood glucose concentrations below 200 mg/dl in the cardiovascular surgery population.^{[3, 4]} The Portland protocol is limited in its applicability to other patient populations such as general surgery or medical populations. In addition, the frequency of hourly blood glucose measurements used in this protocol is a barrier for use in our institution. Other protocols share this same limitation in addition to restrictive target blood glucose ranges, nursing time constraints, and patient discomfort.^[5–11] Published protocols that allow for less-frequent blood glucose measurements that have been shown to be both safe and effective are lacking. Also lacking are protocols that have been studied in a randomized fashion that have produced the same clinical outcomes as seen in the above-mentioned study.^[1]

The primary purpose of our study was to evaluate the safety and effectiveness of blood glucose measurements every 2 hours in maintaining a target blood glucose concentration of 91–130 mg/dl with the use of the intensive insulin infusion protocol at our institution. The secondary objective of this study was to determine if a statistically significant favorable effect on morbidity and mortality was achieved in this sample population.

Methods

Intensive Insulin Infusion Protocol Development

A nurse-managed intensive insulin infusion protocol was developed by a multiprofessional group of physicians, pharmacists, and nursing staff. The protocol was modeled after a previously published study^[12] and restructured to fit the needs of achieving a blood glucose range of 91–130 mg/dl and of measuring blood glucose concentrations every 2 hours (Appendix 1). Blood glucose measurements are performed routinely with capillary finger sticks and the use of the AccuChek Inform system (Roche Diagnostics Corp., Indianapolis, IN). The protocol was first implemented in November 2002 and underwent four pilot trials over 1 year to evaluate for needed improvements in increasing effectiveness and minimizing adverse events. Revisions that were implemented as a result of these pilot trials included instructions for titrating insulin infusion dosages in response to trends of decreasing blood glucose concentrations, improved readability of the protocol to improve adherence to the instructions, and a reduction in frequency of blood glucose measurements to reduce discomfort to patients with the use of finger sticks, particularly those who required the infusion for longer periods of time. Extensive nursing and physician education was provided during these trials and included formal in-services, newsletter articles, and individualized bedside nursing education and coaching. At the time of this evaluation, the protocol was begun by physician preference.

Study Design and Patients

This was a retrospective, observational study in which charts were reviewed for all patients aged 18 years or older who were admitted to the medical ICU and received an insulin infusion. The ICU at St. Vincent Hospital (Indianapolis, IN) is a 32-bed open unit that manages critically ill patients with medical, surgical, or neurologic conditions. At the time of this study, no standard of care was established for physician management of insulin titration rates, initiation of an infusion, or frequency of blood glucose measurements. Patients were excluded from the study if they had a diagnosis of diabetic ketoacidosis or were transferred from another ICU while receiving an insulin infusion. The study was reviewed and approved by the institution's investigational review board.

Data were identified for screening by generation of a report from the pharmacy computer system (RxTFC version 6.1; GE Healthcare Information Technologies, Barrington, IL) for insulin infusion entries. This report was then crossmatched with a list of patient charges for an ICU stay from the hospital financial records (Sunrise Decision Support Manager; Eclipsys Technologies Corp., Boca Raton, FL). Retrospective data were collected for a period of over 1 year (December 1, 2001–November 30, 2002) before initiation of the protocol to establish a control group. Data were then collected retrospectively over a 6-month period (February 1, 2004–August 31, 2004), during which the nurse-managed protocol had been used, to establish the protocol group.

Patient information including age, sex, race, body mass index, weight, a medical history specifically for diabetes mellitus and cancer, reason for ICU admission, length of ICU stay, duration of mechanical ventilation, all-cause mortality, and cause of death as documented in the patient's medical chart were evaluated. In addition, route of nutrition administration (i.e., total parenteral or enteral nutrition), consultation by an endocrinologist, and drug therapy at any time during the ICU stay that may affect blood glucose concentrations (e.g., corticosteroids, thiazide diuretics, atypical antipsychotics) were also included in the evaluation. Frequency of hypoglycemia (defined by a blood glucose concentration < 70 mg/dl), time to achieve blood glucose goal, percentage of blood glucose concentrations within the target range, total number of blood glucose measurements on each infusion day, morning blood glucose concentration (obtained from 3:00–6:00 A.M.) on each infusion day, and day insulin infusion was started with respect to ICU admission were collected in both groups to evaluate the use of the protocol compared with baseline practice.

Statistical Analysis

The SAS version 9.1 software program (SAS Institute, Cary, NC) was used to perform the statistical analysis for the data provided. Student t tests were used to determine the difference in mean blood glucose concentrations, concentrations found within the target blood glucose range, frequency of blood glucose measurements, ICU length of stay, and number of ventilator days. The Kruskal-Wallis test was used to determine the difference in median number of blood glucose measurements performed/patient. The χ² test was used to assess differences in patient demographics as well as differences in nutrition, endocrinology consultation, and drugs used in the sample populations. In addition, a χ² test was performed to assess differences in mortality. A p value of less than 0.05 was considered statistically significant.

Results

Data for 265 patients were screened, and 52 patients were excluded from the evaluation. Reasons for exclusion were the physician managed the infusion (protocol group only, 23 patients), insulin infusion was ordered but never started (13), data were missing (12), patients had diabetic ketoacidosis (3), and patient was managed on another insulin protocol (1). A total of 213 patients met criteria for inclusion in the evaluation: 143 in the control group, 70 in the protocol group. Patient demographics were similar between groups and are highlighted in Table 1 . A statistically significant difference was noted in the type of ICU admissions between groups. An increased percentage of patients admitted to the medical ICU was noted in the protocol group compared with the control group (77% vs 55%, p = 0.001), with the remainder being admitted to the surgical ICU (p = 0.001).

No differences were noted between the groups in route of nutrition administration. Nutrition was administered through the enteral route in 30 (43%) patients in the protocol group and 51 (36%) patients in the control group (p = 0.34). Total parenteral nutrition was administered in 11 patients (16%) in the protocol group and 41 patients (29%) in the control group (p = 0.07). The use of endocrinology consultations was evaluated, but no significant difference was noted between the groups (p = 0.59). Systemic corticosteroid use was more frequent in the protocol group (64%, 45/70 patients) than in the control group (37%, 53/143 patients, p = 0.002). No significant differences were noted between the groups in the use of thiazide diuretics (p = 0.16) or atypical antipsychotics (p = 0.64).

Safety and Effectiveness

Insulin infusions were started on ICU admission day 5.5 (mean) in the protocol group versus day 6 (mean) in the control group. The mean ? SD blood glucose concentration at the start of the infusion was 294 ? 86 mg/dl in the protocol group compared with 348 ? 118 mg/dl in the control group (p = 0.001). The mean ? SD number of blood glucose measurements/patient while receiving the infusion was 54 ? 50 in the protocol group and 37 ? 40 in the control group (p = 0.01). Figure 1 compares the frequency of blood glucose measurements in those patients receiving the infusion for up to 6 days. Among those patients, blood glucose measurements were conducted a mean of every 3 hours in the protocol group and every 4 hours in the control group.

Overall frequency of hypoglycemic events was not significantly different between the groups ( Table 2 ). Although the percentage of patients with hypoglycemic episodes appeared to be higher in the protocol group (43% [30/70]) compared with that in the control group (36% [51/143]), the mean number of hypoglycemic episodes/patient was identical at two hypoglycemic events/patient.

Patients spent a similar amount of time receiving the insulin infusion, with a mean duration of use of 140.4 hours (5.9 days) in the protocol group versus 148.7 hours (6.2 days) in the control group (p = 0.69). The percentage of patients who reached target range (91–130 mg/dl) was also similar between the groups (93% vs 92%, protocol vs control group; Table 2 ). Also, mean ? SD time to reach a target blood glucose was similar between the groups (21 ? 15.3 vs 26.2 ? 29.2 hrs protocol vs control group, p = 0.06). The mean blood glucose concentration while receiving the infusion in the protocol group was lower (170 ? 40 mg/dl) than that in the control group (193 ? 46.3 mg/dl, p < 0.001). In addition, 34% of all blood glucose concentrations in the protocol group were within the target range compared with 23% in control group (relative risk [RR] 1.48, 95% CI 1.38–1.58, p < 0.001). Figures 2 and 3 graphically display the percentage of blood glucose concentrations in the target range and mean morning blood glucose concentrations, respectively, of those patients receiving the infusion for 6 days or less; both graphs depict trends in favor of the protocol group. Once a patient's blood glucose reached target range, 43% of all concentrations in the protocol group remained in the target range compared with 29% of all concentrations in the control group (RR 1.47, 95% CI 1.38–1.56, p < 0.001; Table 2 ).

The median amount of insulin infused on each day was 84 U in the protocol group (range 4–255.4 U) and 91 U in the control group (range 10.5–728 U). The protocol group did show a difference in rates of insulin administration between patients in the surgical ICU and those in the medical ICU (median 39 vs 66 U/day). This difference was not noted in the control group (median 94 vs 91 U/day).

Morbidity and Mortality

The mean ? SD length of ICU stay was lower among patients using the insulin protocol compared with the control group (16.7 ? 10.6 vs 18.4 ? 16 days), but the difference was not statistically significant (p = 0.37). The subset of patients who required more than 14 days of ICU care had a shorter mean length of stay in the protocol group (24 ? 3.9 days) compared with the control group (28.6 ? 12.9 days), a difference that was statistically significant (RR 4.6, 95% CI 0.5–8.72, p = 0.03). In addition, the overall mean days spent with mechanical ventilation was similar between the groups (16.5 ? 9.7 vs 17 ? 15 days, protocol vs control group, p = 0.79). Also, the mean duration of mechanical ventilation was similar in those patients requiring ventilator support for more than 5 days (p = 0.58).

Nineteen patients (27%) died in the protocol group versus 46 patients (32%) in the control group (p = 0.45). Subgroup analysis of those patients alive and in the ICU for more than 5 days showed an overall reduction in mortality in favor of the protocol group (29%) compared with the control group (37%, p = 0.32). Causes of death were recorded ( Table 3 ), and none were noted to be due to insulin therapy.

Discussion

Institutions have strived to develop protocols that achieve tight glycemic control without the research nursing staff as used in the previously mentioned study.^[1] Evaluations of these protocols are variable and present a challenge to clinicians in comparing individual protocols with those published in the literature. For example, one group of authors evaluated the use of a standardized intravenous insulin therapy in 50 patients in their ICU and compared these patients with 50 patients receiving insulin therapy that was titrated by physician discretion.^[10] The investigators reported effectiveness as time spent within the target range of 81–110 mg/dl. The results showed that patients treated in the intervention group with the standardized therapy were within the target range for a mean of 11.5 ? 7.9 hours/day versus 7.1 ? 5 hours/day in the control group (p = 0.0001). Although a statistically significant improvement, the target blood glucose concentration was achieved for less than 50% of the day, challenging the clinical utility of their treatment protocol. Another group of authors found similar results with their protocol.^[7] These investigators evaluated the use of their nomogram and found that only 52% of total blood glucose concentrations in the nomogram group were within the target range of 90–144 mg/dl versus 20% in the control group (p < 0.001). Several other protocols have shown similar results, with a low percentage of blood glucose concentrations within the designated target range.^{[5, 8, 13–15]}

Other investigators, on the other hand, evaluated the use of their insulin infusion nomogram in a sample of patients in the postoperative cardiothoracic surgical ICU.^[11] Their nomogram was modeled off the same published nomogram^[12] as used in the above-mentioned study^[7] and was revised to meet a broader target blood glucose range of 80–150 mg/dl. With the use of hourly blood glucose measurements until target range is achieved for three consecutive measurements and progressing to every 2 hours for four measurements and then every 4 hours if within range, the investigators were able to maintain 61% of all blood glucose concentrations within range as compared with 47% in the control group (p = 0.001). Although this nomogram^[11] has shown the most promising results in maintaining tight blood glucose control, there remains room for improvement in overall blood glucose control.

The study discussed in the introduction^[1] reported mean morning blood glucose concentration as the end point; however, we chose to evaluate the percentage of all blood glucose concentrations in target range to more accurately reflect the consistency of our protocol. Although the percentage of total blood glucose concentrations within the target range was statistically significant, the clinical significance of this difference is arguable. As more protocols are being tested and evaluated in individual institutions, questions remain on what quality assurance measures would most closely define effectiveness of these nomograms and what are the clinically acceptable targets.

After review of the patients who started the insulin protocol, the mean blood glucose concentration at the start of the infusion was 294 ? 86 mg/dl compared with 348 ? 118 mg/dl in our baseline practice. This could be explained by the Hawthorne effect, since the introduction of the protocol was accompanied by extensive education on the importance of blood glucose control to nursing, physician, and pharmacy staff. Therefore, a lower starting blood glucose concentration in the protocol group can be expected. The starting blood glucose concentration may affect how quickly a protocol can achieve the designated target range, which may account for the differences noted in both the previously discussed study^[1] and this evaluation. Although a statistically significant difference was detected in our evaluation, there is little clinical significance to this difference.

The use of a trigger value, as used in other studies, may better depict a protocol's capability of achieving blood glucose control.^{[1, 5, 6, 8, 14, 16]} The type of trigger to implement remains a question for clinicians. The previously discussed study^[1] used a trigger of 110 mg/dl to begin infusions in their study population. Post hoc data analysis published later by the same group demonstrated a reduction in cumulative hazard for inhospital mortality with a blood glucose concentration below 150 mg/dl compared with a blood glucose concentration above this value, suggesting this may be the appropriate starting point for an insulin infusion.^[9] As a result of this data analysis, the Surviving Sepsis Campaign recommends maintaining a blood glucose concentration of less than 150 mg/dl as a more practical approach to blood glucose control outside the research setting.^[17] Some institutions have implemented the use of triggers for initiation of insulin infusions that range from a blood glucose concentration of greater than 110 mg/dl to a blood glucose concentration of greater than 170 mg/dl on multiple, successive measurements.^{[5, 7, 8, 10, 14–16]}

After day 2 of the insulin infusion, a divergence of blood glucose concentrations appears when comparing the protocol group with the control group, with the protocol group showing superiority in percentage of blood glucose concentrations in the target range and mean morning blood glucose concentration. This may be explained by the mean time it took patients to reach the target range in both groups and the larger number of mean and median blood glucose measurements/patient in the protocol group. In addition, the number of cardiac patients seen in the control group may be another reason for this divergence. As seen in the work with the Portland protocol, discontinuation of the insulin infusion may have occurred more frequently on day 3 postoperatively in this large patient population.^{[3, 4]}

The major advantage that our protocol has to offer is a reduced frequency of blood glucose measurements that maintains safety and effectiveness. Despite the protocol's design to achieve a tight blood glucose range, blood glucose measurements were performed at a mean frequency of every 3 hours for patients treated on the protocol. Although there was a statistically significant difference when this was compared with our baseline practice (mean frequency of every 4 hrs), the clinical significance of this difference is negligible. There are limitations to performing more frequent measurements, specifically on an hourly basis. Patient comfort remains a concern with the use of frequent finger sticks. In our earlier experiences, reports of pain, discomfort, and severe bruising at the site of the finger sticks by nonintubated patients and after extubation were observed and worth addressing in our protocol revisions. In addition, more frequent blood glucose concentration checks are likely to increase sleep deprivation, a risk factor for ICU delirium.^[18–21] Perhaps hourly blood glucose measurements would achieve tighter control than what is seen with the use of our protocol, but the expense to the patient is an area that needs to be studied.

There are several situations in which insulin requirements may change, particularly in the critical care setting, which may warrant an increase in the frequency of blood glucose measurement. Compensating for each of these situations increases the complexity of a protocol, which can affect nursing compliance and patient safety, and this may not be a suitable approach for a protocol to practice. For example, changes in corticosteroid management may cause changes in insulin requirements; however, predictability of those changes is based on the pharmacodynamics of the corticosteroid in question. Other examples would include changes in vasopressor requirements, dextrose administration, and temperature curves, all of which may also cause changes in insulin requirements.^[22–25] Due to the wide range of clinical changes that may occur in each patient scenario, approaches to management of these situations limit the practicality of a protocol.

In addition, the pharmacokinetics and dynamics of intravenous regular insulin are unclear since this product is not approved by the United States Food and Drug Administration for intravenous administration. One study performed in an animal model suggests a time to maximum concentration of greater than 1 hour with intravenous administration of human regular insulin.^[26] We observed this to be the case anecdotally in previous pilot trials during the development of this protocol. Based on these findings, increasing the frequency of blood glucose measurements could mislead the clinician in titration of the infusion dosage prematurely. In addition, we believe that a great deal of nursing clinical judgment is warranted in these special circumstances because of the heterogeneous patient population. As a result, we elected to inform the physician in circumstances of abrupt discontinuation of nutrition, and a revision to the protocol was made after this article was accepted for publication, to allow the nurses to increase the frequency of blood glucose measurements if clinically warranted.

The little difference in the amount of insulin administered to our patients during each of the study periods was a surprising finding to the investigators. Unlike what was seen in previous work,^[1] it appeared the patients in our protocol group were able to achieve improved blood glucose control without the use of additional insulin. This may be a result of starting the infusion at a lower blood glucose concentration range and possibly increased time of hyperglycemia before initiation of the infusion in the control group. These data were not evaluated during our study.

Differences were noted in the use of systemic corticosteroids between the study groups, which may have influenced blood glucose control. Increased use of corticosteroids in the protocol group may be explained by the higher number of admissions for respiratory failure. In addition, increasing awareness of relative adrenal insufficiency in the setting of septic shock and treatment with low-dose corticosteroids during the protocol study period may also account for those differences in prescribing patterns when compared with the control group.^{[17, 27]} Despite the higher percentage of patients receiving systemic corticosteroids in the protocol group, the protocol was able to significantly improve on blood glucose control compared with our baseline practice.

The demographic differences between the study groups are explained by changes in admission patterns due to the establishment of a new cardiovascular specialty hospital and an expanding neurosurgical service. Although this change in demographics may limit the application of this particular protocol to the cardiac patient population, our overall improvement in blood glucose control with the use of the protocol demonstrates improvements over our baseline practice.

There are several limitations to this study. The retrospective, observational, chart review design and small sample size are limitations. Evaluation of clinical outcomes was performed in this study, but the small sample limited the ability to truly capture the clinical impact this protocol may have on our heterogeneous patient population. In addition, severity of illness and calorie intake were not taken into account during this evaluation, which may have affected the final results. Although the lower overall mortality rate and ICU length of stay noted in the protocol group may have been associated with improved glycemic control with the use of our standardized protocol, interpretation of our results has additional limitations. Since the time of data collection for the control group, several changes in practice had occurred that might have contributed to shortened ICU length of stay and decreased overall mortality seen in the protocol group. The first is the establishment of an in-house rehabilitation facility that took place during the study period; this allowed for more timely transfers from the ICU. Significant improvements in sedation and pain management strategies at our institution during the study period also may have contributed to the decreased length of stay. Finally, the use of drotrecogin alfa therapy, corticosteroids, and early goal-directed therapy may have affected length of stay in the ICU and overall mortality during the study period.^[27–29] Severity of illness scoring was not collected; therefore, we were unable to determine if these differences existed between the study groups and if this may have affected length of stay.

The primary purpose of this study was to evaluate the protocol's effectiveness in a practical, real-life setting. Selection of patients to be treated with the protocol was not done with consideration of severity of illness, primary diagnosis, corticosteroid prescribing trends, infectious disease process or lack of, diabetes history, obesity, or nutrition requirements, among many other things. Many of the limitations noted in this evaluation are by nature of a retrospective, small, heterogeneous sample population. Since the previously discussed work by another group,^{[1, 9]} we know of no insulin protocol that has been validated for use outside the research setting nor has demonstrated the same outcomes in a heterogeneous patient population. Until a protocol has been validated in a well-designed study, institutions that are able to demonstrate an improvement in practice and maintain safety to the patient will benefit those institutions in the development of their own practice. However, the limitations of this protocol and others must be taken into account when applying a protocol to individualized patient populations.

At the time of acceptance of this publication, the insulin infusion protocol has been in place in the medical, general surgery, cardiovascular surgery, and neurologic ICUs for more than 3 years. Changes to the protocol have since occurred, including the expansion of the target range to 81–130 mg/dl, revisions to starting infusion rates including a customized option for prescribers, and implementation of an insulin infusion trigger. In addition, other local institutions have implemented this protocol and/or a model of this protocol. It is being used in bariatric surgery, trauma, and burn units at other local institutions.

Conclusion

This intensive insulin infusion protocol is safe and improved blood glucose control when compared with our baseline practice. An improvement was seen in overall glycemic control and percentage of blood glucose concentrations within the target range, with no clinically significant differences in hypoglycemic events. In addition, when compared with other published protocols, our insulin protocol displays comparable effectiveness with the use of less-frequent blood glucose measurements.

Characteristic	Control Group (n=143)	Protocol Group (n=70)	p Value
Mean ? SD
Age (yrs)	66 ? 12	60 ? 14	0.13
Body mass index (kg/m2)	30 ? 8	32 ? 8.5	0.11
No. (%) of Patients
Sex			0.11
Male	84 (59)	33 (47)
Female	59 (41)	37 (53)
Race			0.84
Caucasian	122 (85)	59 (84)
African-American	20 (14)	9 (13)
Hispanic	0 (0)	0 (0)
Asian	1 (1)	2 (3)
Medical history
Diabetes mellitus	69 (48)	43 (61)	0.07
Cancer	25 (17)	18 (26)	0.16
Reason for ICU admission
Surgery	65 (46)	16 (23)	0.001
Cardiac (CABG and/or valve replacement)	28	1
Brain (neurologic disease, cerebral trauma)	1	5
Thoracic and/or respiratory insufficiency	1	0
Abdominal (peritonitis)	29	7
Vascular	4	1
Multiple traumas or burns	0	0
Other	2	2
Medical	78 (55)	53 (76)	0.001
Cardiac (myocardial infarction)	6	6
Respiratory distress or insufficiency	23	28
Abdominal, pancreatitis, peritoneal infection	6	1
Neurologic disease or seizures	7	5
Generalized infection	19	7
Other	17	6

ICU = intensive care unit; CABG = coronary artery bypass graft.

 

Blood Glucose Parameter	Control Group (n = 143)	Protocol Group(n = 70)	p Value
Concentration at start of insulin infusion (mg/dl)
Mean ? SD	348 ? 118	294 ? 86	0.001
Median (range)	356 (112–596)	286 (130–540)	—
Concentration while receiving insulin infusion (mg/dl)
Mean ? SD	193 ? 46	170 ? 40	<0.001
Median (range)	184 (128–376)	161 (87–284)	—
Mean no. of hypoglycemic episodes/patienta	2.12	2.13	—
Concentrations relative to target range, no. (%)b
Below	395 (7)	316 (8)	—
Within	1265 (23)	1287 (34)	<0.001
Above	3800 (70)	2159 (57)	—
Parameters after target reached
Patients who reached target concentration, no. (%)	131 (92)	65 (93)	—
Time to reach target concentration (hrs)
Mean ? SD	26.2 ? 29.2	21 ? 15.3	0.06
Median (range)	17.5 (3–214)	17.25 (2–56.5)	—
Concentrations maintained relative to target range, no. (%)
Below	358 (9)	267 (9)	—
Within	1225 (29)	1265 (43)	<0.001
Above	2577 (62)	1347 (48)	—
No. of measurements/patient
Mean ? SD	37 ? 40	54 ? 50	0.01
Median (range)	27 (4–279)	35 (3–291)	0.02

^aHypoglycemic episode defined as blood glucose concentration less than 70 mg/dl.
^bTarget blood glucose concentration range was 91–130 mg/dl.

 

Cause	Control Group (n=46)	Protocol Group (n=19)
Multiple organ failure with proven septic focus	16	0
Multiple organ failure without detectable septic focus	3	1
Severe brain damage	3	2
Acute cardiovascular collapse	12	2
Arrhythmia, ventricular fibrillation or tachycardia	0	0
Infection, systemic inflammatory response syndrome	1	2
Respiratory failure	11	12

 

Appendix: Intensive Insulin Infusion Protocol.

The authors would like to thank the following people for their participation in the development, implementation, and evaluation of this insulin protocol: Mark Smith and Amanda Quebe for their assistance with our study design and use of their statistical knowledge and application; Martha Bailey and Mark Baumgart for their data contribution from the hospital software systems; the ICU nursing staff for their cooperation, compliance, and patience during our protocol development and implementation; Drs. Byron, Cartwright, Holian, Pfeiffer, Vohra, Biondillo, Shapiro, and Strawbridge and the medical residents for their patient care in the ICU.

Presented in part at the Midyear Clinical Meeting of the American Society of Health-System Pharmacists, Orlando, Florida, December 5-9, 2004.

Jennifer A. Quinn, Pharm.D., BCPS, St. Vincent Hospital, Pharmacy Department, 2001 West 86th Street, Indianapolis, IN 46260.