The glucogram: A new quantitative tool for glycemic analysis in the surgical intensive care unit

Abstract

Background: Glycemic control is an important aspect of patient care in the surgical intensive care unit (SICU). This is a pilot study of a novel glycemic analysis tool - the glucogram. We hypothesize that the glucogram may be helpful in quantifying the clinical significance of acute hyperglycemic states (AHS) and in describing glycemic variability (GV) in critically ill patients.

Materials and methods: Serial glucose measurements were analyzed in SICU patients with lengths of stay (LOS) >30 days. Glucose data were formatted into 12-hour epochs and graphically analyzed using stochastic and momentum indicators. Recorded clinical events were classified as major or minor (control). Examples of major events include cardiogenic shock, acute respiratory failure, major hemorrhage, infection/sepsis, etc. Examples of minor (control) events include non-emergent bedside procedures, blood transfusion given to a hemodynamically stable patient, etc. Positive/negative indicator status was then correlated with AHS and associated clinical events. The conjunction of positive indicator/major clinical event or negative indicator/minor clinical event was defined as clinical "match". GV was determined by averaging glucose fluctuations (maximal - minimal value within each 12-hour epoch) over time. In addition, event-specific glucose excursion (ESGE) associated with each positive indicator/AHS match (final minus initial value for each occurrence) was calculated. Descriptive statistics, sensitivity/specificity determination, and student's t-test were used in data analysis.

Results: Glycemic and clinical data were reviewed for 11 patients (mean SICU LOS 74.5 days; 7 men/4 women; mean age 54.9 years; APACHE II of 17.7 ± 6.44; mortality 36%). A total of 4354 glucose data points (1254 epochs) were analyzed. There were 354 major clinical events and 93 minor (control) events. The glucogram identified AHS/indicator/clinical event "matches" with overall sensitivity of 84% and specificity of 65%. We noted that while the mean GV was greater for non-survivors than for survivors (19.3 mg/dL vs. 10.3 mg/dL, P = 0.02), there was no difference in mean ESGE between survivors (154.7) and non-survivors (160.8, P = 0.67).

Conclusions: The glucogram was able to quantify the correlation between AHS and major clinical events with a sensitivity of 84% and a specificity of 65%. In addition, mean GV was nearly two times higher for non-survivors. The glucogram may be useful both clinically (i.e., in the electronic ICU or other "early warning" systems) and as a research tool (i.e., in model development and standardization). Results of this study provide a foundation for further, larger-scale, multi-parametric, prospective evaluations of the glucogram.

Figure 1

Schematic representation of the open-high-low-close (O-H-L-C) system. All glucose values within each 12-hour period (or epoch) were “compressed” into this simplifi ed, structured graphical form, with the only values of importance being the O-H-L-C data. This format is subsequently utilized to construct secondary graphs, including moving averages and specialized indicators

Figure 2

Glucogram example. Red squares represent 24 major clinical events correctly matching with corresponding indicator spikes. Green squares represent seven minor clinical events that correctly correlated with lack of indicator spike. In terms of mismatches, there were three instances of a major event not correlating with an indicator spike (empty squares) and four indicator spikes incorrectly correlating with minor events or no apparent event (gray squares). Of note, the number of events associated with each occurrence/spike is listed within the corresponding square. The uppermost window shows the momentum indicator (MoIR). In this case, positive MoIR spikes were considered to have values of 200% or greater (areas shaded in yellow). The middle window shows the stochastic indicator (StIR). In this case, values of 60 or above were considered to represent positive indicator spikes (areas shaded in yellow). The bottom window shows the glucose levels represented in the O-H-L-C format

Figure 3

Composite graphs of glycemic variability (GV) during the last 100 pre-discharge 12-hour epochs. Discharge was defi ned as either ICU mortality or discharge alive from the ICU. Dashed red line represents GV for non-survivors and black solid line represents GV for survivors. Composite raw GV data are shown in (A) while composite 10-epoch moving average for GV is shown in (B). Note that the 10-epoch GV moving averages interact only once during the entire 100-epoch period (black circle). This fi nding may be pertinent to daily patient care because the GV, represented as a moving average, could serve as a clinical “barometer”. GV is defi ned as the maximal–minimal glucose value for each epoch, with glucose levels expressed in mg/dL