Objective: To measure the response time of health professionals before sound alarm activation and the implications for patient safety. Method: This is a quantitative and observational research conducted in an Adult Intensive Care Unit of a teaching hospital. Three researchers conducted non-participant observations for seven hours. Data collection occurred simultaneously in 20 beds during the morning shift. When listening the alarm activation, the researchers turned on the stopwatches and recorded the motive, the response time and the professional conduct. During collection, the unit had 90% of beds occupied and teams were complete. Result: We verified that from the 103 equipment activated, 66.03% of alarms fatigued. Nursing was the professional category that most provided care (31.06%) and the multi-parameter monitor was the device that alarmed the most (66.09%). Conclusion: Results corroborate the absence or delay of the response of teams, suggesting that relevant alarms might have been underestimated, compromising patient safety.