Can J Diabetes 38 (2014) 139e143

Contents lists available at ScienceDirect

Canadian Journal of Diabetes

journalhomepage:
www.canadianjournalofdiabetes.com

Original Research

Assessment of Quality of Glycemic Control in Intensive Care Patients Treated

with an Insulin Infusion at a Teaching Hospital
a, * a b
Lyne Gauthier BScPhm, MSc , Jessica Ferguson PharmD, MSc , Anne-Isabelle Dubé PharmD, MSc , Patrick Viet-
a c d
Quoc Nguyen BSCPhm, MSc , Marie-France Beauchesne BScPhm, MSc, PharmD , Jean-Marie Boutin MD, PhD
aDépartement de Pharmacie, Centre hospitalier de l’Université de Montréal, Montreal, Quebec, Canada
b Pharmacy Department, McGill University Health Centre, Montreal, Quebec, Canada
c Département de Pharmacie, Centre hospitalier de l’Université de Sherbrooke, Sherbrooke, Quebec; and Centre de recherche clinique Etienne-Le Bel, Sherbrooke and Faculté de
Pharmacie, Université de Montréal, Montreal, Quebec, Canada
d Service d’Endocrinologie, Centre hospitalier de l’Université de Montréal, Montreal, Quebec, Canada
article info
Article history:
Received 3 December 2013
Received in revised form 22
January 2014
Accepted 26 January 2014
Keywords: critical
care glycemic
control
intravenous insulin
Mots clés : soins
critiques
maîtrise glycémique
insuline intraveineuse
abstract
Objective: To describe the quality of glycemic control in patients in intensive care units (ICUs) treated with an intravenous
(IV) insulin infusion at a teaching hospital.
Method: This retrospective study included patients admitted to the ICU and treated with an IV insulin infusion for at least 12 h
between August 1 and November 30, 2011. Medical charts were reviewed. The primary quality indicator for glycemic control
was the mean percent of blood glucose values per patient in the 6.1 to 8 mmol/L target range.
Results: A total of 351 patients were included; 61.5% of subjects had no known diabetes. Admissions were mainly for surgery
(61.3%). The mean Acute Physiology and Chronic Health Evaluation (APACHE) II score was 16.8 7.3. The mean percent of
blood glucose values per patient in the 6.1 to 8 mmol/L range was 35% for all subjects and 26.2% for patients with diabetes. If
a target of 6.1 to 10 mmol/L was considered, those values became 63% and 54.6%. At least 1 episode of hyperglycemia (>10
mmol/L), hypoglycemia (<4 mmol/L) or severe hypoglycemia (<2.2 mmol/L) was documented in 68%, 9% and 1% of
subjects, respectively. Glycemic variability (SD) was 1.9 mmol/L, and the median hyperglycemic index was 0.77 (interquartile
[IQ]: 0.24 to 1.63).
Conclusion: The quality of glycemic control in patients in the ICU at our hospital needs to be improved. A new computerized
IV insulin protocol is currently being tested.
2014 Canadian Diabetes Association
résumé
Objectif : Décrire la qualité de la maîtrise glycémique de patients hospitalisés à l ’unité de soins intensifs (USI) d’un hôpital
d’enseignement et traités par perfusion intraveineuse (IV) d’insuline.
Méthode : Cette étude rétrospective incluait les patients admis à l’USI et traités par perfusion IV d’insuline durant au moins 12
er
h entre le 1 août et le 30 novembre 2011. Les dossiers médicaux ont été passés en revue. Le principal indicateur de qualité de
la maîtrise glycémique était le pourcentage moyen des valeurs de la glycémie par patient selon une fourchette cible de 6,1 à 8
mmol/l.
Résultats : Un total de 351 patients ont été inclus; 61,5 % des sujets n ’avaient aucun diabète connu. Les admissions relevaient
principalement de chirurgies (61,3 %). Le score moyen selon l’APACHE II (Acute Physiology and Chronic Health Evaluation
II) était de 16,8 7,3. Le pourcentage moyen des valeurs de la glycémie par patient qui étaient situées dans la fourchette de 6,1 à
8 mmol/l était de 35 % pour l’ensemble des sujets et de 26,2 % pour les patients souffrant de diabète. Si une cible de 6,1 à 10
mmol/l était prise en considération, ces valeurs grimpaient à 63 % et à 54,6 %. Au moins un épisode d’hyperglycémie (> 10
mmol/l), d’hypoglycémie (< 4 mmol/l) ou d’hypoglycémie grave (< 2,2 mmol/l) a été documenté chez 68 %, 9 % et 1 % des
sujets, respectivement. La variabilité de la glycémie (écart-type) était de 1,9 mmol/l, et l’indice hyperglycémique médian était
de 0,77 (IQ : 0,24 à 1,63).
 * Address for correspondence: Lyne Gauthier, BScPhm, MSc, Département de Pharmacie,
Centre hospitalier de l’Université de Montréal-Hôtel-Dieu, 3840 St-Urbain, Montreal, Quebec
H2W 1T8, Canada.
E-mail address: lyne.gauthier.chum@ssss.gouv.qc.ca.
1499-2671/$ e see front matter 2014 Canadian Diabetes Association
http://dx.doi.org/10.1016/j.jcjd.2014.01.012
140 L. Gauthier et al. / Can J Diabetes 38 (2014) 139e143
Conclusion : La qualité de la maîtrise glycémique chez les patients de l’USI de notre hôpital doit être améliorée. Un nouveau
protocole informatisé sur la perfusion IV d’insuline est actuellement testé.
Introduction
Glycemic control is an important aspect of care for critically ill patients.
Hyperglycemia is associated with an increased risk for mortality, especially in
patients not known to have diabetes (1e5). Optimal glycemic targets for
patients in the intensive care unit (ICU) setting are still a matter of debate.
The most recent recom-mendation from professional associations is to
maintain blood glucose (BG) levels between 7.8 and 10 mmol/L for the
majority of patients in the ICU. A lower target (6.1 to 7.8 mmol/L) may be
appropriate in selected patients if feasible without significant hy-poglycemia
(6,7).
An intravenous (IV) insulin infusion is recommended to treat
hyperglycemia in patients in the ICU (6,7). An IV insulin protocol with
proven efficacy and safety should be used (7). There is no single ideal
protocol. Some key elements of a safe and effective protocol are having a
clear indication of when to start and when to stop the infusion, clear glycemic
targets, infusion rates determined by the rate of change of glycemia (not only
by the last glucose measurement) and frequent BG measurements (7e9). The
protocol should be easy to prescribe, easy to follow and associated with a
hypoglycemia protocol, as well as a transition to a subcutaneous insulin
protocol (7e10).
Many of the desired characteristics associated with an optimal IV insulin
protocol are not implemented at our hospital. The aim of this study was to
evaluate the quality of glycemic control in patients in the ICU who were
treated with an IV insulin infusion at our teaching hospital.
Methods
The Centre hospitalier de l’Université de Montréal is a teaching hospital
that includes 3 distinct geographic sites (Hôtel-Dieu, Notre-Dame and St-
Luc), with a total of 52 mixed ICU beds. This study was approved by the
Ethics Committee of the Centre hospi-talier de l’Université de Montréal. No
informed consent was required because all the data were anonymized and
collected retrospectively. Adult patients admitted to the ICU between August
1 and November 30, 2011, who received an IV insulin infusion for at least 12
consecutive hours, were eligible. Subjects had to have at least 3 measured and
documented BG values. Exclusion criteria were pregnancy, ketoacidosis and a
hyper-osmolar hyperglycemic state. All the patients admitted to the ICU
during the study period were found by the Meditech (Westwood, MA)
software used by the hospital pharmacy department at the time. The
computerized medical chart (OACIS; Telus Health; Longueuil, QC) for each
patient was consulted to see whether IV insulin had been used and whether
the other inclusion and exclusion criteria had been met. The practice for
glucose control during the study period was the use of a static IV insulin
protocol, where the rate of infusion is determined by the actual BG. The
major characteristics of this protocol were a BG target range of 6.1 to 8
mmol/L and a usual initial insulin concentration of 0.1 units/mL that was
modified based on BG values; no hypoglycemia or tran-sition protocol were
embedded. The 3 sites did not use exactly the same protocol, and there were
many variations in application. The BG value at which to start the insulin
infusion and guidance on frequency of BG testing were physician dependent.
2014 Canadian Diabetes Association
The main objective was to describe the quality of glycemic control. The
primary quality indicator used was the mean propor-tion per patient of BG in
the 6.1 to 8 mmol/L target range of the protocol and in the broader 6.1 to 10
mmol/L target range. Other quality indicators were measured to provide a
better description of glycemic control: mean ( SD) BG; time to reach target
(6.1 to 8 mmol/L or 6.1 to 10 mmol/L); proportion of patients with at least 1
hyperglycemia (>10 mmol/L), 1 severe hyperglycemia (>15 mmol/L), 1
hypoglycemia (<4 mmol/L) or 1 severe hypogly-cemia (<2.2 mmol/L).
Glycemic variability was calculated by the standard deviation of the mean BG
level. The hyperglycemic index is defined as the area under the glucose curve
above the upper limit of normal divided by the length of stay (11), calculated
for the duration of the infusion, with 8 mmol/L as the upper limit. The
secondary objective was to describe manipulations related to the IV insulin
infusion: mean BG at which the infusion was started, mean daily number of
BG measurements and mean daily number of concentration and rate changes.
Glycemic control was studied for the duration of the infusion or a maximum
of 120 h for each subject in the study.
BG is usually measured with capillary samples obtained by finger stick;
arterial blood is sometimes used. The point-of-care meter used in our hospital
is the Accu-Check Inform II (Roche Diagnostics, Indianapolis, IN). A quality
program is in place for calibration of the meters. Nurses enter BG values,
times of mea-surements, concentrations of insulin and infusion rates on a
daily ICU chart.
Demographic characteristics of patients and all the data were found in the
computerized medical charts (OACIS). Data on age, sex, weight, APACHE II
and Sequential Organ Failure Assessment (SOFA) scores, admission
diagnosis, presence of pre-existing diabetes, use of hyperglycemic medication
and nutrition status were collected. Data were analyzed with Statistical
Package for the Social Sciences (SPSS; IBM, Chicago, IL) software. Missing
data were simply ignored, except for the APACHE II and SOFA scores, in
which missing data were replaced by normal values. Means and confidence
in-tervals were calculated by the Student bilateral t test. Differences in
2
proportions were analyzed by the c test. Median delay to reach target was
evaluated by a Kaplan-Meier analysis. Based on the hypothesis that 65% of
BG values would be in the 6.1 to 8 mmol/L target range, a bilateral Fisher
exact test was used to determine that a sample size of 348 patients would be
needed to estimate that 65% 7% (with alpha ¼ 0.05 and 80% statistical
power) were on target. The hypothesis that 65% of BG values would be on
target was based on results of other studies (12e17), although results are
difficult to compare: BG targets, protocols used and even ways of calculating
proportions vary a great deal. Post hoc analysis of sub-groups was performed
to evaluate the impact of certain patient characteristics on glycemic control.
Results
A total of 351 subjects, evenly distributed across the 3 sites, were
included from 675 charts reviewed. A total of 322 patients were excluded
because of an insulin infusion of fewer than 12 h, and 1 patient was excluded
due to pregnancy. Patients’ characteristics are detailed in Table 1.
The
study population comprised a majority of men (58.4%) with a
mean age of 61.7 years. No history of diabetes was noted for
61.5% of the subjects. There were 61.3% surgical cases,
L. Gauthier et al. / Can J Diabetes 38 (2014) 139e143 141
Table 1 (CI 95%, 10.4 to 11). Patients had a mean of 8.5 (CI 95%, 8.3 to 8.7) BG
Demographic and medical characteristics of subjects measurements done per day and a daily mean of 4 (CI 95% 3.8 to 4.2)
Number of patients 351 changes in the rate of the insulin infusion. A change in infusion concentration
Mean age in years 61.7 13.9 was needed in 124 (35%) subjects, with a mean of 1 (CI 95% 0.8 to 1.1)
Male gender 205 (58.4)
Mean weight in kg 81.9 21.3 change per day. Post hoc analysis of subgroups did not show statistically
Diabetes No history 216 (61.5) significant differences in glycemic control based on sex, type of admission
Type 1 5 (1.5) (medical or surgical) or treatment with corticosteroids. The proportion of
Type 2 130 (37)
blood glucose values on target (6.1 to 8 mmol/L) was lower for patients with
Type of admission Medical 136 (38.7)
Surgical 215 (61.3) diabetes (26.2%) than for patients without diabetes (40.5%) (p<0.001). Also,
APACHE II score 16.8 7.3 the mean BG was 8.7 mmol/L for the whole population studied, with a
SOFA score 5.8 3.4 difference between patients who had or did not have dia-betes (9.5 mmol/L
Admission diagnosis Cardiac surgery 104 (29.6)
and 8.2 mmol/L, respectively; p¼0.001).
Other surgery 96 (27.4)
Neurology 39 (11.1)
Pulmonary 27 (7.7)
Hemodynamic/cardiology 19 (5.4) Discussion
Microbiology 19 (5.4)
Gastroenterology 17 (4.8)
Transplantation 16 (4.6) Leuven I and NICE-SUGAR are 2 landmark studies on glycemic control
Burn 8 (2.3) in critically ill patients (12,18). It is interesting to compare their population
Nephrology 2 (0.6) with ours, with obvious limitations. Our patients were more severely ill than
Hematology 1 (0.3)
the patients in Leuven I and less severely ill than the subjects in NICE-
Other 3 (0.9)
Medication know to Vasopressor 132 (37.6)
SUGAR, based on the APACHE II score. Diabetic patients constituted 38.5%
cause hyperglycemia Corticosteroid 111 (31.6) of our popu-lation compared to 13% in Leuven I and 20% in NICE-SUGAR.
Calcineurin inhibitor 29 (8.3)
Pentamidine 1 (0.3) Our study showed a low proportion of BG values in the target range (6.1
Nutrition Fasting 243 (69.5)
to 8 mmol/L) in our patients in the ICU. Although the chosen target is 6.1 to 8
Tube feeding 84 (23.9)
Oral 15 (4.3) mmol/L at our hospital, controversy sur-rounds those values. Some clinicians
Parenteral 9 (2.6) might prefer to aim for a broader target range (6.1 to 10 mmol/L). When that
interval is considered, the proportion of on-target BG values is obviously
Results are presented as mean SD or number (%).
better but still not optimal. The low percentage of BG values in the chosen
target range may also possibly be related to weak compliance with the
with about 30% of patients admitted post-cardiac surgery (nearly half of the protocol, which was not evaluated in our study. Clearly, there is a need to
surgeries). A mean APACHE II score of 16.8 typified the population. reassess the protocol and its application.

Table 2 details the results of various indicators of quality of glycemic

control. Patients received IV insulin infusions for a mean of 57.3 h and
obtained mean BG levels of 8.7 mmol/L. The mean proportion of BG values
in the 6.1 to 8 mmol/L target was 35%, which is significantly different from
the 65% value proposed in our hypothesis (p<0.001). For a target range of 6.1
to 10 mmol/L, 63.3% of BG values were on target. At least 1 episode of
hyperglycemia (>10 mmol/L), severe hyperglycemia (>15 mmol/L),
hypoglycemia (<4 mmol/L) and severe hypoglycemia (<2.2 mmol/L) was
seen in 68%, 18%, 9% and 1% of patients, respectively. Glycemic variability
was 1.9 mmol/L. A median hyperglycemic index of 0.77 mmol/L was
calculated. A median delay of 10 h was needed to reach target (6.1 to 8
mmol/L); that value became 2 h for a target range of 6.1 to 10 mmol/L.
Results for the secondary objectives showed that the mean blood glucose
at which the infusion was started was 10.7 mmol/L
The mean BG of our population was similar to the values obtained in
control groups for Leuven I (mean morning BG of 8.5 mmol/L) and NICE-
SUGAR (mean morning BG of 8 mmol/L). However, the occurrence of
severe hypoglycemia in these populations, respectively 0.8% and 0.5% of
patients, was slightly lower than in our population (18,12). Hypoglycemia has
been associated with higher mortality rates in patients in the ICU (19,20).
Even mild hypoglycemia has deleterious effects (19).
Glycemic variability has been found to be a significant predictor of
mortality in critically ill patients, independently of severity of illness (21e23).
Standard deviation is not the best way to measure variability, but it is simple
and it is used frequently (22). In our analysis, glycemic variability was higher
in patients with diabetes (2.1 vs. 1.9 mmol/L; p¼0.01). In the conventional
groups in the Leuven 1 and NICE-SUGAR studies, glycemic variability was
1.8

Table 2
Description of glycemic control

Indicator Results
Duration of the insulin infusion (h) 57.3 (53.1e61.5)*
Mean proportion of blood glucose in target per patient 6.1e8 mmol/L 35.0 (32.7e37.3)*
6.1e10 mmol/L 63.3 (60.8e65.8)
Mean blood glucose (mmol/L) 8.7 (8.5e8.9)*
Glycemic variability (SD- mmol/L) 1.9 (1.8e2.1)*
Median delay to reach target (h) 6.1e8 mmol/L 10
6.1e10 mmol/L 2
Median hyperglycemic index (mmol/L) 0.77 (0.24e1.63)y
Patients with at least 1 episode (%) Severe hypoglycemia (<2.2 mmol/L) 3 (1)
Hypoglycemia (<4 mmol/L) 32 (9)
Hyperglycemia (>10 mmol/L) 269 (68)
Severe hyperglycemia (>15 mmol/L) 70 (18)
* Results presented as mean and 95% CI.
y Median (interquartile range).
142 L. Gauthier et al. / Can J Diabetes 38 (2014) 139e143
mmol/L and 1.4 mmol/L, respectively (calculated as the SD of mean morning
BG) (18,12). Krinsley (22) reported glycemic variability in a cohort of
critically ill patients. Patients were grouped into sub-populations according to
mean glucose levels. Ranges of standard deviations were grouped by quartiles
for each of those sub-populations. Our population, based on its mean BG
level, would be in the second quartile of SD. Although this result seems
acceptable, its clinical significance is unclear.
A median time to reach the target of 10 h seems long, even if there is no
defined optimal time. A comparison of various protocols showed that time to
target varied widely, from 2 h to more than 48 h, and computerized protocols
tended to perform better (10).
The hyperglycemic index quantifies hyperglycemia (11). The goal is to
aim for a value as close to 0 as possible. The hyperglycemic index, when
calculated with 6 mmol/L as the upper limit of normal, has been associated
with mortality (11). Subjects with an index higher than 1 mmol/L had 2 to 3
times the risk for mortality compared to subjects with an index less than 1
mmol/L. The hy-perglycemic index we calculated seems relatively good. It is
important to note, however, that the result would have been more accurate if
more BG measurements had been taken.
In this study, a mean of 8.5 BG values were measured per patient per day.
Although the optimal frequency of BG monitoring in patients in the ICU
receiving IV insulin infusions has not been established, the recommendation
is to measure BG every 1 to 2 h and more frequently after an episode of
hypoglycemia (every 15 m after treatment until blood glucose >4 mmol/L)
(6e8). An interval of 4 h is used in some protocols when BG values are on
target, but this practice can be associated with unrecognized hypoglycemia
(8). At our hospital, checking BG every 4 h is an extremely common practice,
and it may have contributed to some of the hypoglycemia episodes.
Point-of-care meters are used at our centre, as in the majority of hospitals,
although the accuracy of this method is judged to not be optimal (8,9). Precise
information about sampling sites could not be retrieved from the medical
charts, but finger-stick capillary sampling is usually done, with occasional
arterial sampling.
The practice of changing the concentration of the insulin infusion
increases the risk for errors and adds to the workload. The initial
concentration used at our hospital is nearly always 0.1 units/mL and is
increased as needed on the basis of BG values. Based on an analysis of insulin
errors at our hospital over the past 3 years, insulin infusion ranked second or
third on the list of insulin errors, mainly concentration and rate errors.
Standardized con-centration is recommended, usually 1 unit/mL, and a
change of practice will be proposed at our hospital.
This study gives a good picture of the quality of glycemic control and use
of insulin infusion in the 3 critical care units in our hospital. The fact that the
study was done retrospectively prevented people from changing their usual
practices. Data were retrieved for pa-tients treated in 2011. Since the study, a
unique IV insulin protocol based on practice at each site has been used in our
ICUs. The BG target is still 6.1 to 8 mmol/L. Prescribers still modify
elements frequently, so we can say that the practice is the same as it was.
There is a need to standardize the use of IV insulin at our hospital. A project
for computerized insulin protocol has been proposed, with the aim of
standardizing practice and improving glycemic control. The quality of
glycemic control has been described by numerous and varied indicators that
will facilitate comparison with the new computerized protocol under
development.
We were using a retrospective analysis, so data had to be extracted from
medical charts that had some missing or unclear values, which is a limitation
of this study. The fact that BG values and times of measurements were not
available in computerized format (no downloads of this information from
meters) may have led to inaccuracies. Data were retrieved from paper charts
filled in by nurses, and times of BG testing had to be approximated.
No analysis of mortality and morbidity outcomes was per-formed because
that was not the purpose of the study. This article describes the situation at
our hospital, but these results can be used by others to compare their practices
and confirm that glycemic control in the ICU remains a challenge.
Conclusion
The quality of glycemic control in the critical care units at our teaching
hospital was not optimal, especially in patients with dia-betes. The proportion
of BG values in the intended target range was relatively low. Obviously,
results may be found to be more accept-able if the higher blood target
proposed by some diabetes associ-ations is considered. Whatever the chosen
target, there is room for improvement. Many elements of the IV insulin
protocol used could be modified. There is a need to standardize the use of IV
insulin in our ICUs. A project for computerized insulin protocol has been
proposed; it aims to improve quality of glycemic control and the quality of
care for our patients.
Author Disclosures
LG has received honoraria, speaker fees and consulting fees from Eli Lilly
and NovoNordisk and speaker fees from AstraZeneca, Bayer, Boehringer
Ingelheim, Bristol-Myers Squibb, and Sanofi Aventis. JF and AID have no
conflicts of interest to declare. PVQN has worked as a pharmaceutical
representative at Ferrer Internacionale. MFB has received unrestricted grants
from NovoNordisk, Sanofi Aventis and Eli Lilly. JMB has received speaker
fees from AstraZeneca, Boehringer Ingelheim, Bristol-Myers Squibb, Eli
Lilly, NovoNordisk, and Sanofi Aventis and was a member of advisory boards
for those companies, as well as for Janssen.
Author Contributions
LG designed the study and wrote and edited the manuscript; JF, AID and
PVQN designed the study, researched the data and reviewed the manuscript.
MFB and JMB collaborated on the design of the study and reviewed the
manuscript.
References
1. Dungan KM, Braithwaite SS, Preiser JC. Stress hyperglycaemia. Lancet 2009;
373:1798e807.
2. Umpierrez GE, Isaacs SD, Bazargan N, et al. Hyperglycemia: an independent marker of in-
hospital mortality in patients with undiagnosed diabetes. J Clin Endocrinol Metab
2002;87:978e82.
3. Krinsley JS. Association between hyperglycemia and increased hospital mor- tality in a
heterogeneous population of critically ill patients. Mayo Clin Proc 2003;78:1471e8.
4. Christiansen C, Toft P, Jorgensen HS, et al. Hyperglycaemia and mortality in critically ill
patients: a prospective study. Intensive Care Med 2004;30: 1685e8.
5. Falciglia M, Freyberg RW, Almenoff PL, et al. Hyperglycemia-related mortality in
critically ill patients varies with admission diagnosis. Crit Care Med 2009;37: 3001e9.
6. Houlden R, Capes S, Clement M, et al. Canadian Diabetes Association 2013 Clinical
Practice Guidelines for the Prevention and Management of Diabetes in Canada: in-hospital
management of diabetes. Can J Diabetes 2013;37(Suppl 1): S77e81.
7. Moghissi ES, Korytkowski MT, DiNardo M, et al. American Association of Clinical
Endocrinologists and American Diabetes Association consensus state- ment on inpatient
glycemic control. Diabetes Care 2009;32:1119e31.
8. Jacobi J, Bircher N, Krinsley J, et al. Guidelines for the use of an insulin infusion for the
management of hyperglycemia in critically ill patients. Crit Care Med 2012;40:3251e76.
9. Ichai C, Preiser JC. International recommendations for glucose control in adult nondiabetic
critically ill patients. Crit Care 2010;14:R166.
10. Krikorian A, Faramarz IB, Moghissi ES, et al. Comparisons of different insulin infusion
protocols: a review of recent literature. Curr Opin Clin Nutr Metabol Care
2010;13:198e204.
L. Gauthier et al. / Can J Diabetes 38 (2014) 139e143 143
11. Vogelzang M, van der Horst IC, Nijsten MW. Hyperglycaemic index as a tool to assess
glucose control: a retrospective study. Crit Care (London, England) 2004; 8:R122e7.
12. Finfer S, Chittock DR, Su SY, et al. Intensive versus conventional glucose control in
critically ill patients. N Engl J Med 2009;360:1283e97.
13. Zimmerman CR, Mlynarek ME, Jordan JA, et al. An insulin infusion protocol in critically
ill cardiothoracic surgery patients. Ann Pharmacother 2004;38: 1123e9.
14. Orford N, Stow P, Green D, et al. Safety and feasibility of an insulin adjustment protocol
to maintain blood glucose concentrations within a narrow range in critically ill patients in
an Australian level III adult intensive care unit. Crit Care Resusc 2004;6:92e8.
15. Krinsley JS. Effect of an intensive glucose management protocol on the mortality of
critically ill adult patients. Mayo Clin Proc 2004;79:992e1000.
16. Goldberg PA, Siegel MD, Sherwin RS, et al. Implementation of a safe and effective insulin
infusion protocol in a medical intensive care unit. Diabetes Care 2004;27:461e7.
17. Cyrus RM, Szumita PM, Greenwood BC, et al. Evaluation of compliance with a paper-
based, multiplication-factor, intravenous insulin protocol. Ann Phar- macother
2009;43:1413e8.
18. van den Berghe G, Wouters P, Weekers F, et al. Intensive insulin therapy in the critically
ill patients. N Engl J Med 2001;345:1359e67.
19. Egi M, Bellomo R, Stachowski E, et al. Hypoglycemia and outcome in critically ill
patients. Mayo Clin Proc 2010;85:217e24.
20. Hermanides J, Bosman RJ, Vriesendorp TM, et al. Hypoglycemia is associated with
intensive care unit mortality. Crit Care Med 2010;38:1430e4.
21. Egi M, Bellomo R, Stachowski E, et al. Variability of blood glucose concentration and
short-term mortality in critically ill patients. Anesthesiology 2006;105: 244e52.
22. Krinsley JS. Glycemic variability: a strong independent predictor of mortality in critically
ill patients. Crit Care Med 2008;36:3008e13.
23. Hermanides J, Vriesendorp TM, Bosman RJ, et al. Glucose variability is associated with
intensive care unit mortality. Crit Care Med 2010;38: 838e42.