BACKGROUND: Moderate sedation is routinely performed in patients undergoing minor therapeutic and diagnostic procedures outside the operating room. The level of sedation is often monitored by sedation nurses using clinical criteria, such as sedation scores. The Bispectral Index (BIS) is derived from changes in the electroencephalograph profile that may provide an objective measure of the level of sedation. In this prospective observational study, we investigated whether using BIS values to guide sedative drug administration influences the level of sedation and the incidence of adverse events compared with using Ramsay sedation scale (RSS) only in nurse-administered moderate sedation. We hypothesized that both depth of sedation and the incidence of adverse events related to oversedation would decrease when sedation nurses used BIS values to help guide sedative drug administration. METHODS: Sedation care was provided by trained sedation nurses under the supervision of a physician performing the procedure. The sedation regimen was initiated with IV midazolam 1 to 2 mg and fentanyl 50 mcg or hydromorphone 0.2 mg. Additional small boluses of midazolam, fentanyl, or hydromorphone were administered to maintain an RSS of 2 to 3 (cooperative, oriented, and responding to verbal command). Propofol was not used. Information including patient demographics, type of procedure, medication administered, RSS, and rates of adverse events was recorded by the sedation nurses for each patient on a computer-readable form. The study was divided into 3 phases. In phase 1 (baseline, 6 months' duration), baseline data on sedation practice were prospectively collected. There was no change from standard of care for all patients except that each patient had a BIS sensor attached, but the monitor was covered and nurses were blinded to the BIS values. In phase 2 (training, 3 months), the sedation nurses received comprehensive education on the use of BIS to guide sedative drug administration, pharmacology of commonly used drugs, and methods for rescue from oversedation. The recommended BIS range for moderate sedation was 75 to 90. Adequate training of all sedation nurses on the use of BIS was documented. In phase 3 (implementation, 6 months), the BIS values were used to guide drug administration. RESULTS: Data were available on 1766 patients (999 and 767 patients in phases 1 and 3, respectively). Most of the procedures were colonoscopies, upper gastrointestinal endoscopies, examinations under anesthesia, endoscopic retrograde cholangiopancreatography, and central venous access catheter placements. No differences in the demographics between the 2 groups were observed. The RSS was inversely associated with the BIS value, r = -0.16 (95% confidence interval, -0.19 to -0.12; P < 0.00001). An RSS of 2 to 3 was maintained in 94% of patients in phase 1 and 96% of patients in phase 3 The mean (±SD) BIS values were 80.9 ± 6.8 in phase 1 and 80.4 ± 6.5 in phase 3. The number of sedation-related adverse events was lower in our sample when BIS was used, with an odds ratio of 0.41 (95% confidence interval, 0.28-0.62; P < 0.0001), and patients with restlessness had a lower BIS value than those without this symptom (P < 0.0001). No serious adverse events or deaths were reported. CONCLUSIONS: Nurse-administered moderate sedation using midazolam and fentanyl was usually associated with appropriate levels of sedation as assessed by both the RSS and BIS with an overall low incidence of adverse events. The use of BIS did not change the mean level of sedation significantly, although the number of sedation-related adverse events appears to be lower when BIS was used.