Current methods used to evaluate depth of sedation in Intensive Care Unit (ICU) patients are subjective.   Presently, the Ramsey Sedation Score (RSS) is the primary method used to determine sedation level (Habibi and Coursin 1996). This technique employs a numerical scale by which an observer can rank the patient's response to verbal commands. The observer evaluates how aware the patients are by calling out their names and prodding the patients.   The patient is then scored using the RSS system.  The inefficacy of such a system is threefold. First, intubated and unconscious patients cannot be effectively evaluated; because they cannot respond to the observer's command, they will be given a score of 0 to 1, and they may be unnecessarily sedated. Second, RSS is subject to inter-observer variability and is not an objective measure of sedation.  Third, RSS can only differentiate into two categories of sedation: responder, score of 4 to 6, and non-responder, score of 0 to 3.  Although there are seven levels one can give the patient, it is difficult to assess the difference between 0, 1, 2, or 3 scores as well as between 4, 5, or 6 scores.  Thus, RSS is not precise enough to ensure that the patient will get the ideal amount of sedative.

A more objective measure of sedation can be provided using an electroencephalography (EEG) machine that measures the patients brain activity. In the following study, we used Bispectral Index (BIS) analysis of the EEG (Aspect 1992).  The EEG machine computes a multivariate discriminate analysis called BIS (Aspect 1992). BIS analysis quantitatively measures consistency between phase and power of EEG waves and computes a single number, BIS. BIS can portray the dynamic structure of EEG that simple linear analysis cannot.   Therefore, BIS analysis can depict subtle changes in brain activity (Aspect 1992). Higher doses of anesthetics cause slower, more synchronized EEG patterns that are picked up by BIS analysis.  BIS can help one detect cerebral state changes when patients are under anesthesia (Aspect 1992).  Previous studies have shown BIS analysis to be a good indicator of depth of anesthesia in healthy volunteers (Glass et al. 1997; Lui et al. 1997).  BIS was found to correlate with level of responsiveness and predict the loss of consciousness in volunteers receiving propofol, midazolam, isoflurane, or alfentanil (Lui et al. 1997). Also, BIS was found to correlate with the observer's assessment on an alertness and sedation scale (Glass et al. 1997).  However, we wanted to determine if BIS correlates with RSS, and if it is a good indicator of sedation levels in ICU patients.

Our data suggest that BIS correlates well with RSS measurements of response to verbal command.  Hence, BIS can be used to monitor and assess sedation level in ICU patients.  BIS is a more objective, precise tool of assessing levels of anesthesia than RSS alone.


Materials and Methods

Subjects and setting:
Following approval by the Institutional Review Board (IRB) of our proposed study, we tested BIS efficacy in monitoring and assessing sedation levels of 30 consenting male patients, aged 46 to 90, in the surgical ICU at the Long Beach Veterans Administration Hospital. Some subjects were on a ventilator; many were receiving some type of sedative or analgesic medication.  This was documented, but not controlled for in the study.


An Aspect A-1000 EEG monitor was used to monitor patients' brain electrical activity in real time.  A four-channel EEG recording was performed using ZipPrep electrodes placed in a fronto-temporal montage for 6 to 8 h. Other parameters monitored were EKG, pulse oximeter, heart rate, and blood pressure.  BIS was computed by the EEG and recorded every half hour.  Patients' response to verbal commands was measured every half hour as well.  The research assistant assessed patients' response to verbal commands by calling out the patient's name, and then shaking and prodding if the patient was asleep. The patient was given a score based on the RSS (Table 1).  The patients with lower scores (0 to 3) were pooled into one group called the "non-responders." Patients with higher scores (4 to 6) were pooled into a second group termed the "responders."  All of the data was documented on a computer, including any new medications, treatments, or muscle movements the patient underwent during the study period.  This was done to account for any drastic changes in EEG patterns.

Table 1
Ramsay Sedation Scale (Adapted from Habibi and Coursin 1996)

table1.gif (6737 bytes)


The EEG data recorded during the study period was downloaded to a computer that synchronized all parameters. EEG parameters that were processed include BIS, spectral edge frequency (SEF), median power frequency (MPF), absolute power of all frequencies (delta, theta, alpha, beta), and low EMG (Table 2).

Statistical analysis included Levene's test for equality of variance and independent samples t-test for equality of means.  A p< 0.05 was considered statistically significant.  Linear regression analysis for correlation of RSS with BIS was performed as well.  Data from the left and

back.gif (221 bytes) next


Page 2


Susan Khera - Bispectral Analysis of EEG Waves Provides... [1] [2] [3] [4]