Frequency of Clinical Alarms in Intensive Care Units and Nurses' Sensitivity to Them

An Observational Study

Burcu Ceylan, PhD, RN; Leyla Baran, PhD, RN; Ülkü Yapucu Güneş, PhD, RN


Am J Crit Care. 2021;30(3):186-192. 

In This Article



This observational study was conducted from October 2016 through March 2017.

Setting and Participants

The research took place in 2 medical (cardiology, internal medicine) and 2 surgical (anesthesiology and reanimation, neurosurgery) ICUs at a university hospital in Izmir, Turkey. The bed to nurse ratio at these ICUs was 10:10 in the cardiology ICU, 14:11 in the internal medicine ICU, 13:26 in the anesthesiology and reanimation unit, and 14:16 in the neurosurgical ICU. These units were included in the study because of their intensive use of alarms and the fact that they are open units rather than made up of private rooms.

Clinical alarms are extremely important to patient safety, but can threaten patient safety if not managed properly.

Sample selection was performed through the purposive sampling method. All nurses who provided routine care to critically ill patients were eligible for inclusion in the study, with the exception of nurses with less than 6 months of experience in the ICU.

Each ICU engaged in patient monitoring had the same types of alarm-sounding devices. These included bedside monitors that measure parameters such as electrocardiographic activity, heart rate, respiration rate, oxygen saturation, noninvasive blood pressure, and invasive blood pressure; pulse oximeters; mechanical ventilators; fluid infusion pumps; and enteral feeding pumps. Bedside monitors, pulse oximeters, and mechanical ventilators emit both critical and warning alarms, whereas infusion and feeding pumps give off just warning alarms. All of the patients were being monitored. The patients in the anesthesiology and reanimation unit were connected to all of the devices. The internal medicine unit was the unit with the fewest patients connected to a mechanical ventilator, followed by the cardiology and the brain and neurosurgical units. Over the course of the study, a total of 20 nurses (5 from each unit) were observed, with each observation period lasting 1 hour. Nurses who had worked in these ICUs for 6 months or more and agreed to participate in the study were enrolled. The sample size was based on research that used similar study designs.[8,13]

Data Collection Procedures

After we selected the nurses to be recruited and obtained their consent to participate, we obtained permission from each unit supervisor to observe each nurse in the ICU during their shift. We reviewed the working hours of the nurses and made arrangements so that they could complete the 4-hour observation requirement. The observations took place during the daytime (8 AM to 4 PM) and nighttime (4 PM to 8 AM) shifts; the total observation time was 80 hours.

Clinical alarm frequency, alarm type (valid, false, technical), and nurses' response to the alarm were observed in 4 intensive care units.

Two researchers who had experience as ICU nurses (B.C., L. B.) collected the data. They took their places in each unit, situated where they could see the nurse being observed proceed to take care of each patient. A manual counter was used to count alarm sounds. Each time an alarm sounded, the researchers recorded the medical device generating the alarm and the cause of the alarm. The nurse's response to the alarm was also recorded. A chronometer was started as soon as the alarm sounded and was stopped when the nurse responded, and the nurse's response time was noted.

Data Collection Instruments

Data were collected by means of a structured observation form. The researcher was prompted to record the alarm category, the number of alerts sounded, the reason for the alarm, the nurse's response to the alarm, and the duration of the alarm's activation. Alarms were categorized as valid, false, or technical. An alarm that sounded as a result of an event that did not indicate actual physiological distress was a "false alarm" (ie, indicating a deviation from normal values because of sensor displacement). An alarm that sounded as a result of actual physiological events was a "valid alarm." An alarm that sounded as a result of a technical error was considered a "technical alarm" (eg, depleted charge, device probe not sensing, disconnection of probe from patient).[7,14]

A measurement tool devised by Priscilla K. Gazarian[8] was used to describe the nurses' reactions to the alarms. This tool features the following 5-point Likert-type scale: 0 = no response, 1 = visual checking of the patient, 2 = checking the alarm device but no assessment of the patient, 3 = nursing intervention (assessing patient and/or device, changing position, etc), 4 = change in care plan (ie, physician informed, a new order drawn up, and tests arranged for). The form was tested for validation of Turkish linguistic equivalence and for reliability by independent observers. It was translated into Turkish after permission was obtained from the tool's author. Four nursing instructors and 1 English-language specialist translated the form from English into Turkish. We compared 5 translated versions to develop the form in Turkish. Then, the initial translation into Turkish was translated back into English by an English-language specialist who had not seen the original English text. The form's statements, translated from Turkish into English, were compared with the statements in the original English text and the necessary revisions were made. To test the form's reliability, 2 researchers simultaneously observed 3 nurses who were not a part of the study for a period of 2 hours. To determine interrater agreement between the observers, intraclass correlation and κ coefficients were calculated. After the observation period, the researchers reviewed the observation form for clarity and agreed on its validity. According to the interrater agreement analysis, the κ value was 0.87 and the intraclass correlation coefficient was 0.94.

Data Analysis

Data analysis was performed using IBM SPSS Statistics, version 24. Categorical variables were represented with numbers and percentages, and continuous variables with means and SDs.

The Kruskal-Wallis H test was used to determine the differences among the ICUs in terms of the total mean number of alarms sounding over the course of 1 hour. The Mann-Whitney U test was used to determine the differences according to shifts. The level of statistical significance was set at P less than .05.

Ethical Considerations

Written permission to conduct the study was obtained from the Ege University Faculty of Nursing Scientific Ethics Committee (No. 2016–259) and from the hospital where the study was carried out. The nurses recruited into the study were provided with information about the study and their oral consent to participate was obtained. Also, permission to use the measuring tool in the study was obtained from its author.