Blood Glucose Control in Critically Ill Adults and Children^*: A Survey on Stated Practice FREE TO VIEW

The best blood glucose target and management strategy for critically ill patients are presently unclear. The evidence supporting the benefits of tight glucose control (80 to 110 mg/dL [4.4 to 6.1 mmol/L]) in critically ill patients is questioned by those concerned with hypoglycemic events associated with a lower glucose target.¹²³ Furthermore, the applicability of the results of the study by Van den Berghe et al² (decreased mortality in critically ill surgical patients with tight glucose control) to children or critically ill adult medical patients remains in question.¹⁴ Despite the lack of a common algorithm to manage blood glucose levels with insulin and a concern for increasing hypoglycemia, many intensivists have adopted a target blood glucose level of 80 to 110 mg/dL (4.4 to 6.1 mmol/L). The overall practice pattern of North American intensivists with regard to the definition of hypoglycemia and hyperglycemia, the choice of blood glucose target range, and the guidelines for insulin infusion are undefined.

An understanding of clinician practice patterns and beliefs is essential for future widespread improvement of blood glucose control algorithms. This survey explores these practice patterns and beliefs.

The study protocol for this survey was approved by the Scientific Committees of the Reengineering Clinical Research in Critical Care Network, the National Institutes of Health/National Heart, Lung, and Blood Institute ARDS Network (ARDSnet) and the Pediatric Acute Lung Injury and Sepsis Investigators (PALISI) Network. This survey was conducted with the approval of the institutional review board of Sainte-Justine Hospital, which approved the waiving of informed consent. Respondents were advised that consent to participate in this study was implied if they answered and returned the questionnaire.

The questionnaire was sent to intensivists practicing in academic ICUs in the United States and Canada. Members of the following groups were asked to participate: Reengineering Critical Care Clinical Research Network; the ARDSnet; and the PALISI Network. We aimed to obtain at least one questionnaire per participating ICU. The medical director of each ICU that was contacted was instructed to enlist one registered nurse, one fellow, and also at least one other attending intensivist to complete the questionnaire.

The term blood glucose control refers to active interventions by health-care providers to modify blood glucose concentrations.

The first draft of the questionnaire was developed at an informal meeting held in Salt Lake City, UT, on March 16, 2005. Participants included four intensive care physicians, one fellow in training, and a medical informatics specialist. This first draft was sent for review to four adult intensivists, six pediatric intensivists, one fellow in training, and two experts in informatics. These collaborators independently reviewed the text, and added comments and new questions. A second draft was sent to the same experts. The item-generation process was considered to be complete once no new items were added. Three iterations were required to complete the item-generation portion of this survey.

Questions that were considered to be irrelevant, or redundant, by the majority of experts (> 50%) were deleted. Based on this process, 14 questions were deleted, and 45 questions were retained. The 45-question survey included the number of questions in the following categories: clinician/institution characteristics (6 questions); ICU description (9 questions); hyperglycemia (15 questions); hypoglycemia (5 questions); blood glucose measurement (7 questions); and the computer experience of respondents (3 questions). The full questionnaire is available on the following Web site: http://www.coordinatornih.org/survey/BAAGlucoseSurvey.pdf.

The self-administered questionnaire used to perform this survey was formatted following established quality criteria.⁵⁶

Pretesting of the questionnaire to ensure clarity, relevance, and completeness was performed using semistructured interviews with two independent pediatric intensivists and two independent adult intensivists.

Four clinicians working in the adult ICU along with six clinicians working in the pediatric ICU rated the potential ability of the questionnaire to discriminate among respondents. These clinicians also rated the clarity, utility, face validity, content validity, and redundancy of the questionnaire.⁵ They used a 7-point Likert scale ranging from 1 (complete disagreement) to 7 (complete agreement). A response of 5, 6, or 7 was considered to be satisfactory, and 76% of respondents had scores of 5, 6, or 7 on all questions (2 of 14 reviewers accounted for almost all scores of < 5). We considered this evaluation to be satisfactory.

The questionnaire was posted on the Utah Clinical Trial Toolbox Website, and e-mail invitations to complete the questionnaire were sent to all listed critical care units ^{(Appendix 1)}. The mailing agenda followed the recommendation of Woodward et al.⁷ Up to six personalized e-mail reminders were sent to participants with incomplete replies. We sent a series of e-mail reminders (at 2, 4, and 6 weeks) after detecting a 35% response rate. After a 2-month response-free period, the survey was considered to be complete. (The Utah Clinical Trials Toolbox was created by Dean Sorenson [^{see Appendix 2}].)

The responding clinicians directly entered all data into a Web-based survey relational database.

Descriptive statistics of dichotomous or ordinal variables are proportions; those of continuous variables are the mean ± SD or the odds ratio and its 95% confidence interval. Data were analyzed using a statistical software package (SPSS for Windows, version 13.0.1; SPSS Inc; Chicago, IL). We determined the significance of differences between adult and pediatric ICU clinicians with χ² analysis (for categoric variables) and independent Student t test (for continuous variables). The results were reviewed by The Scientific Committees of the Reengineering Clinical Research in Critical Care Network, the National Institutes of Health/National Heart, Lung, and Blood Institute ARDSnet, and the PALISI Network.

We generated and validated the questionnaire from May to August 2005. Fifty of the 52 institutions returned at least one questionnaire (96% response rate for all institutions). Seventeen of 19 adult sites (90%) and 33 pediatric site (100%) returned at least one questionnaire. The overall clinician response rate was 58% (163 of 282 physicians responded) [^{Appendix 1}].

Based on ICU site, 36% of respondents (58 of 163 respondents) were adult ICUs, 64% were pediatric ICUs (104 of 163 respondents), and 1 respondent was unknown (Table 1 ). The response rates of adult practitioners (58 of 101 responses) and pediatric practitioners (104 of 181 responses) were both 57%.

Twenty-six adult respondents and 42 pediatric respondents (3 of which were unknown), totaling 71 respondents, answered the questions targeted to medical directors (Table 2 ). This is the only area of the questionnaire that was not completed by all respondents. Thirty-nine percent of the 71 respondents provided patient severity score data. The stated mortality rates were 15.1% for adult ICUs and 4% for pediatric ICUs. The average number of patients per nurse was 1.8 (range, one to three patients per nurse).

The average threshold at which respondents defined hyperglycemia or initiated insulin therapy was 145 mg/dL [8.0 mmol/L]). A significant difference existed, however, between adult ICU clinicians (120 mg/dL [6.7 mmol/L]) and pediatric ICU clinicians (150 mg/dL [8.3 mmol/L]; t = 5.41; p < 0.001) [Fig 1 ].

The factors (1, irrelevant; 5, very important) associated with high blood glucose concentrations were as follows: exogenous administration of glucocorticoids and the patient’s endogenous stress response to their critical illness (mean score, 4.6); the presence of type I or II diabetes (mean score, 4.2); and the presence of multiple organ dysfunction syndrome and treatment with catecholamines (mean score, 4.0). Most clinicians also perceived the administration of exogenous IV glucose (mean score, 3.6) as a contributing factor (Table 3 ).

The perceived percentage of ICU patients who persistently had blood glucose higher than 120 mg/dL (6.66 mmol/L) during their ICU stay was significantly different between adult and pediatric clinicians (χ² = 24.4; p < 0.001) [Fig 2 ]. Seventy-seven participants (47.3%) stated that they used a guideline or protocol to manage hyperglycemia. Blood glucose target ranges were highly variable across respondents, and significantly differed between adult and pediatric sites (χ² = 13.6; p = 0.009) [Table 4 ].

The disease states that were perceived to contribute to the initiation or modulation of blood glucose control (1, unlikely; 5, very likely) were sepsis, severe sepsis, multiple organ system dysfunction, shock, and ARDS. All disease states had mean rankings of ≥ 4.5, suggesting that blood glucose control is likely to be attempted in many critically ill medical patients. Physicians were less likely to initiate blood glucose control during postoperative care following cardiac surgery (mean score, 3.9). The (multiple answers allowed) most important determinants of insulin dose were patient diagnosis (72.4%), duration of hyperglycemia (58.9%), previous insulin infusion rate (79.1%), current insulin infusion rate (84.0%), and/or previous (48.5%) and current (74.8%) administration of glucose or other nutrients.

Ninety-three percent of all clinicians stated that the titration of the insulin infusion rate should depend on the rate of change of blood glucose, but they reported a wide range of time periods between measured glucose levels and the change in insulin that was considered to be reasonable. Sixty-four percent of adult ICU clinicians and 33% of pediatric ICU clinicians stated the initial insulin infusion rate was based on the current blood glucose level. Another 33% of pediatric clinicians reported using a starting dose of 0.1 U/kg/h.

There were no significant differences between adult and pediatric clinicians in terms of the blood glucose level used to define hypoglycemia (Fig 3 ). Most clinicians defined hypoglycemia at a concentration of ≤ 60 mg/dL (3.33 mmol/L). Estimates of the incidence of hypoglycemia (using their stated thresholds) varied (Fig 4 ). There were significant differences in their choice of treatment for hypoglycemia.

Fifty-nine percent of adult ICU clinicians and 85% of pediatric ICU clinicians thought that hypoglycemia was more dangerous than hyperglycemia. The perceived risk factors for hypoglycemia (1, not important; 5, very important) were as follows: delay in insulin titration (mean score, 4.2); stopping the glucose drip (mean score, 4.06); stopping enteral feeding (mean score, 3.96); liver failure (mean score, 3.89); and adrenal insufficiency (mean score, 3.42).

This survey addresses clinician’s perceptions of hyperglycemia, hypoglycemia, and blood glucose control with insulin, and is the first survey to report on the blood glucose practices of both pediatric and adult intensivists. To our knowledge, only one survey⁸ has been published regarding the practice pattern of intensivists with respect to blood glucose control; this survey involved Canadian intensivists working with critically ill adults. McMullin et al⁸ noted that their respondents defined hypoglycemia as a glucose concentration of < 4 mmol/L (60 mg/dL or 3.3 mmol/L in this survey), and hyperglycemia as > 10 mmol/L (this survey: adults, 120 mg/dL or 6.7 mmol/L; children, 150 mg/dL or 8.3 mmol/L). The data from these two surveys suggest that no consensus definition of hyperglycemia or hypoglycemia exists for critically ill patients.

This survey highlights the gap between what practitioners perceive and what is found in the literature. The range of glucose targets and definitions of hyperglycemia and hypoglycemia vary widely. The decision algorithms for insulin initiation, titration, and hypoglycemia prevention also vary greatly. Clinicians underestimate the frequency of hyperglycemia and hypoglycemia in their ICUs. In this survey, most adult intensivists perceived the incidence rate of hyperglycemia (blood glucose concentration, > 120 mg/dL [6.7 mmol/L]) in critically ill patients as being between 25% and 50%, while the incidence rate observed at the bedside by Finney et al⁹ was roughly 70%. Similarly, the majority of pediatric intensivists using the same definition have reported the incidence rate of hyperglycemia as < 25%, while Srinivasan et al⁴ (blood glucose concentration, > 126 mg/dL) and Faustino and Apkon¹⁰ (blood glucose concentration, > 150 mg/dL) reported incidence rates of 86% and 45%, respectively. Physician perceptions formulate the foundation for their practice patterns. The existence of this gap and the variation in practice pattern suggest a need for better knowledge transfer and additional trials on blood glucose control.

Understanding the perceptions and practice patterns of clinicians is critical for designing the clinical environment in which future trials will be conducted. The ARDSnet, for example, used a postal survey to determine a common definition of ARDS to use in their landmark low tidal volume study.¹¹¹² Similarly, Hebert et al¹³ and Laverdiere et al¹⁴ conducted surveys of blood transfusion practices before they undertook two large randomized clinical trials on blood transfusions in adult and pediatric ICUs. These surveys defined the threshold hemoglobin concentrations for the experimental and control groups of future trials.¹⁵¹⁶ The results of this survey could help to design algorithms and define blood glucose target ranges for future trials in critically ill patients.

Most respondents to this survey perceived hyperglycemia and hypoglycemia as being dangerous. Both groups of respondents defined hypoglycemia with a median level of blood glucose level ≤ 60 mg/dL (3.3 mmol/L), a value that is higher than that commonly accepted in the literature (≤ 40 mg/dL [2.2 mmol/L]). Less than 50% of subjects reported using a guideline or protocol reflecting the variability present in blood glucose control practices. Despite the concern for hypoglycemia and the evidence for the association of tight glucose control with increased rates of hypoglycemia,¹⁷ the majority of adult clinicians (73%) and almost half of pediatric clinicians (43%) chose a blood glucose target of 80 to 110 mg/dL (4.4 to 6.1 mmol/L) in their ICU. The percentage of pediatric clinicians targeting this range is surprising given that no prospective evidence supports the control of blood glucose in critically ill children and that many think hypoglycemia poses a greater risk than hyperglycemia. Clinicians have unresolved questions about both the risks and benefits of tight glucose control (4.4 to 6.1 mmol/L).

Hyperglycemia is indeed common in both critically ill adults and children,¹⁰¹⁸ and studies¹⁴⁹¹⁰ have shown a relationship between hyperglycemia and increased mortality. In vitro data^241018192021 have suggested that sustained hyperglycemia is detrimental to critically ill patients because it disturbs many important metabolic functions. Despite the possible benefits of tight glycemic control, there is an increasing concern among practitioners about causing hypoglycemia.²²²³ A recent blood glucose control trial²² in adult medical ICU patients (the Volume Substitution and Insulin Therapy in Severe Sepsis [or VISEP] trial) was stopped prematurely in part because of high hypoglycemia rates. There is little evidence to confirm the harm induced by a transient episode of hypoglycemia. The results of this survey emphasize, however, the growing concern on the part of clinicians about causing hypoglycemia. Although there is considerable evidence to support blood glucose control, the issue is far from settled, particularity for children.

Issues common to survey reports limited this study. Our data rely on self-reporting and perceived prevalence, both of which incorporate a reporting bias. Our sample includes 50 institutions, and the clinician response rate was 58%. This represents a 96% institutional response rate and should have been high enough to gain a picture of caregivers’ perceptions regarding hyperglycemia, hypoglycemia, and blood glucose control in the ICU. Our survey is directed at a wide sample of North American practitioners who were affiliated with research networks. This sample may not be reflective of the practice patterns of other non-university-affiliated institutions.

In summary, caregivers working in the ICU for adults and for children underestimate the frequency of hyperglycemia in their practice when compared to the literature and define hypoglycemia at a cutoff higher than that accepted in the literature. Although most subjects believe that tight blood glucose control should be considered for at least some critically ill patients, practice patterns vary widely. The definitions of hyperglycemia and the ideal blood glucose target for IV insulin titration differ between adult and pediatric practitioners. Pediatric intensivists are more concerned about the possible consequences of hypoglycemia than adult intensivists. Regardless of their affiliation with adult or pediatric populations, it seems that the clinicians who were questioned in this survey believed that the issue of blood glucose control in critically ill patients has many unanswered questions.

Given the controversy present in the literature and the varying opinions among clinicians about glycemic control, more large multicenter trials using the same treatment algorithm are required to answer the remaining questions. Documenting clinician practice patterns is a critical step in generating a standardized protocol with widespread acceptance. A common algorithm with “testable” methodology for glucose control will decrease practice variability and is mandatory to ascertain whether the mortality and morbidity benefits of tight glucose control outweigh the potential harm of increasing hypoglycemia.

Invitations to complete the questionnaire were sent by e-mail to the following critical care units: Baystate Medical Center, Springfield, MA; Centre Mere-Enfant du Centre Hospitalier Universitaire de Québec, Sainte Foy, QC, Canada; BC Children’s and Women’s Hospital, Vancouver, BC, Canada; Children’s Hospital and Regional Medical Center, Seattle, WA; Children’s Hospital and Research Center at Oakland, Oakland, CA; Children’s Hospital at Dartmouth, Hanover, NH; Children’s Hospital of Austin, Austin, TX; Children’s Hospital of Central California, Madera, CA; Children’s Hospital of Denver, Denver, CO; Children’s Hospital of Los Angeles, Los Angeles, CA; Children’s Hospital of Michigan, Detroit, MI; Children’s Hospital of Philadelphia, Philadelphia, PA; Children’s Hospital of Pittsburgh, Pittsburgh, PA; Children’s Hospital of Wisconsin, Milwaukee, WI; Children’s Hospitals and Clinics Minneapolis/St. Paul, MN; Cornell Medical Center, New York, NY; Denver General, Denver, CO; Doernbecher Children’s Hospital, Portland, OR; Duke Children’s Hospital, Durham, NC; Duke University, Durham, NC; Earl K. Long Medical Center, Baton Rouge, LA; Fresno Medical Center, Fresno, CA; Hackensack University Medical Center, Hackensack, NJ; Harborview Medical Center, Seattle, WA; Hopital Sainte-Justine, Montréal, QC, Canada; Johns Hopkins Hospital, Baltimore, MD; Kosair Children’s Hospital, Louisville, KY; LDS Hospital, Salt Lake City, UT; Louisiana State University, Baton Rouge, LA; Lucille Salter Packard Children’s Hospital, Palo Alto, CA; Massachusetts General Hospital, Boston, MA; Mayo Eugenia Litta Children’s Hospital, Rochester, MN; Miami Children’s Hospital, Miami, FL; H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL; Montreal Children’s Hospital, Montreal, QC, Canada; Moses H. Cone Memorial Hospital, Greensboro, NC; Penn State Children’s Hospital, Hershey, PA; University of Pittsburgh Medical Center, Pittsburgh, PA; St. Louis Children’s Hospital, St. Louis, MO; University of Chicago, Chicago, IL; University of Virginia, Charlottesville, VA; University Hospitals Cleveland, Cleveland, OH; University of Alberta Hospital, Edmonton, AB, Canada; University of Florida College of Medicine, Gainesville, FL; University of Maryland, Baltimore, MD; University of Michigan, Ann Arbor, MI; University of Michigan Health System; University of Virginia Children’s Medical Center, Charlottesville, VA; Vanderbilt University, Nashville, TN; Wake Forest University, Winston-Salem, NC; and Yale University School of Medicine, New Haven, CT.

Phillipe Jouvet, MD, Université de Montréal, Montréal, QC, Canada; Neal Thomas, MD, Penn State University, College Park, PA; Taylor Thompson, MD, Harvard University, Boston, MA; Arno Zaritsky, MD, University of Florida, Gainesville, FL; Gordon R. Bernard, MD, Vanderbilt University, Nashville, TN; Peter Cox, MD, University of Toronto, Toronto, ON, Canada; and Paul C. Hébert, MD, University of Ottawa, Ottawa, ON, Canada.

Dean Sorenson, PhD, University of Utah, Salt Lake City, UT.

Catherine Oldmixon, RN, MA, General Hospital, Boston, MA.

Reengineering Clinical Research in Critical Care Network, the PALISI Network, and the National Institutes of Health/National Heart, Lung, and Blood Institute ARDS Network.

Nancy Ringwood, RN (Boston, MA); and Kathy Sward (Salt Lake City, UT).

Catherine Ann Farrell, Karen Harrington, Géraldine Pettersen, Anna Proietti, François Proulx, Miriam Santschi, and Nathalie Thomas, all from St. Justine Hospital (Montreal, QC, Canada) contributed to the validation of the questionnaire.