Alarm Management in an ICU Environment Carola van Pul 1,2, Wouter Dijkman 3, Heidi van de Mortel 4, Jarno van den Bogaart 5, Thilo Mohns 4,5, Peter Andriessen 4 1 Maxima Medical Center, Department of Clinical Physics, Veldhoven, The Netherlands c.vanpul@mmc.nl 2 Eindhoven University of Technology, School of Medical Physics and Engineering, Eindhoven, The Netherlands 3 Maxima Medical Center, Department of Intensive Care, Veldhoven, The Netherlands w.dijkman@mmc.nl 4 Maxima Medical Center, Department of Neonatology, Veldhoven, The Netherlands (h.vandemortel, t.mohns, p.andriessen)@mmc.nl 5 Maxima Medical Center, Department of Medical and Information Technology, Veldhoven, The Netherlands j.vandenbogaart@mmc.nl Abstract. In the intensive care unit (ICU) environment, a large number of alarms is generated. The exact number varies per unit and per hospital. To improve the alarm chain, an alarm management system was implemented that distributed alarms to a handheld. The incidence and type of alarming (alarm pressure) in this distributed alarm system in an adult ICU and a neonatal ICU (NICU) in a teaching hospital were investigated. Methods to improve the alarm chain and reduce alarm pressure are proposed. Keywords. Patient monitoring Alarm pressure Intensive Care Environment Introduction To care for critically ill patients, various types of equipment (e.g. patient monitor, ventilator and infusion pumps) are used in order to treat and monitor the patient s condition continuously [1]. Alarms are used for each device to alert the healthcare provider when a variable exceeds a predefined threshold or when there is a sensor/device problem. However, the rate of alarms is increasing [2], and the rising alarm pressure is considered to be a safety issue by the Joint Commission and the Emergency Care Research Institute [3,4]. The large number of alarms may create an unsafe and noisy environment, and may lead to the occurrence of alarm fatigue of health care providers, since less than 20% of all alarms is clinically relevant [5,6]. The rate of alarms per patient differs per unit and depends on the chosen settings/equipment and type of patient. Our objective is to evaluate the incidence and type of alarms in a Neonatal Intensive Care Unit (NICU) and an intensive care unit (ICU), in order to analyze alarm pressure, for future improvement of the alarm chain.
Materials and Methods The NICU in Máxima Medical Center (MMC) is a 18-bed level III NICU with an admission rate of approximately 380 newborns per year. The NICU comprises of 9 single, 5 twin rooms and one triplet room. In a twin or triplet room the possibility exists to care for mother and child together. The adult ICU at MMC is a 14-bed level II ICU with 530 adult admissions per year. This unit has 8 single rooms and 3 double rooms. Both ICU environments are equipped with a patient monitor (NICU with a Philips patient monitor and ICU with GE patient monitor), ventilator and infusion pump arrays and at the NICU also an incubator (Fig. 1). To distribute alarms from the room through a central location to a handheld of a nurse, an alarm management system has been implemented (Fig. 2). The primary alarm chain (orange) consists of the patient monitor in the room, inter-bed communication functionality that allows identification of alarms in another room, and a central monitor. This primary chain connects to the server (Ascom) in the secondary chain, for the Philips monitor this is achieved via an Emergin server, whereas for the GE monitoring system, a direct connection to the Ascom server is made. The secondary alarm chain (green) consists of the Ascom server and wireless network that transfer alarms to the handheld of the nurse. All components in the alarm chain are monitored to check continuously that the complete system is working. A user group (both in ICU and NICU) determined for the patient monitor the alarm limits and the urgency of the alarms; only urgent alarms are sent to the central system. In Table 1, the urgent alarms and corresponding alarm limits are shown. The system is configured in a way that back-up alarms are sent to a buddy nurse if alarms are not confirmed within 45 s (monitor) or 60s (PCS/DCS system). We call this the repeated alarm. From 18 th of February to 18 th of July 2014, alarm-logs of the central system (Ascom, Sweden) and all alarms of devices connected via PCS/DCS or patient monitor were stored and analyzed using an in-house developed software program (Mathematica). Note that only the selected (filtered) alarms from the patient monitor are sent to the handhelds in the distributed system and only these alarms are stored in this logging. Table 1. Alarm limits of alarms sent to the handheld, for NICU (left) and ICU (right). GA=gestational age, SaO2=oxygen saturation, ABP= arterial blood pressure. High and low refer to the high and low limits of the physiologic variable. Philips monitoring at NICU GE monitoring at ICU Red (urgent) alarms NICU GA <26 wk GA 26-36 wk GA 37 wk Crisis alarms ICU Heart rate high (bpm) 230 230 230 Asystole Heart rate low (bpm) 80 80 60 Ventricular fibrillation SaO 2 low (%) 80 80 80 Ventricular tachycardia Apnea time (s) 20 20 20 Mean ABP high (mmhg) 55 75 75 Mean ABP low (mmhg) 19 25 31
Fig. 1 NICU room (left) and ICU room (right) Fig. 2. Alarm chain: patient call system (PCS) and device-call system (DCS) both send alarms to the handheld only as a simple indicator. The patient monitor is connected to the primary chain (central monitor, interbed-communication) and via Medical Alarming (MA) system, an urgency-based selection of alarms is sent to the handheld, containing information about alarm type and parameters. Results In a 5-month period, an average of 12,8 patients per day stayed at NICU and generated ~228k alarms at the handheld (146k monitoring alarms, 43k ventilator, 17k infusion), equivalent to 5 alarms per hour per patient. The average gestational age of the patients was 31,3 weeks with an average birthweight of 1758gr. At ICU, an aver-
age of 10 patients (on average 77% ventilated) generated 138k PCS/DCS alarms and 7k monitor crisis alarms, the total corresponding to 4 alarms per patient per hour at the handheld of the nurse. The difference in relative alarm distribution during the day is shown in Fig.3: for NICU, the largest source of alarms is the patient monitor (oxygen desaturation followed by bradycardia alarms). For ICU, the largest source of alarms is the ventilator, followed by infusion. For NICU, 12% of all PCS/DCS alarms were repeated to the buddy, whereas for monitor alarms, only 6% were repeated. In the ICU, the alarm goes first to the first responsible nurse, the repeated alarm to both first responsible nurse and the buddy (but in our analysis we cannot distinguish between these two in the logging). The second repetition of the alarms goes to the complete group of handhelds ( third escalation ). The alarms are repeated until the alarm condition is solved. The third escalation alarms can be distinguished from the other alarms in our analysis. On average 10% of all ICU alarms was sent to the third escalation (none of them patient monitor alarms). For NICU no distinct patterns as a function of daytime is observed, whereas at ICU, the morning nursing moment (arrow) can be distinguished. Fig. 3. Number of alarms per patient per hour of the day at NICU (left) and ICU (right). Discussion Both adult and neonatal ICU are equipped with highly sophisticated (monitor) devices, resulting in 4 (IC) to 5 (NICU) alarms per patient per hour distributed to the handhelds. This difference is caused by the choices of which alarms to send to the handheld: in the adult ICU only critical monitor alarms are distributed to the handheld, as nurses continuously view the central monitor or interbedcommunication. In contrast, at the NICU, nurses are mainly in the patient rooms, using the interbed-communication functionality of the patient monitor and they also use the handheld when outside the patient rooms, therefore the user group chose to send more alarms to the handheld. In the ICU, ventilator alarms cannot be filtered based on urgency and all are sent to the handheld, while in NICU, only urgent ventilator alarms are sent (via patient monitor).
The number of repeated alarms gives an indication of the experienced usefulness of the type of alarm. The reaction to the patient call (PCS) is clearly less than to the device call system (DCS). Reaction to monitor alarms is best, only 6% goes to the buddy at NICU and at ICU, no patient monitor alarm went to third escalation level. Clear daytime variation in alarms is observed at the ICU, mainly caused by the morning nursing moment. This is due to the fact that silencing alarms during patient care is not common and even not possible for all devices. At NICU, nurses are used to silencing monitor alarms during patient care. During the night, at NICU no patient call alarms are observed, since parents (who use this call system) stay in the rooms during daytime, but they prefer to stay in another room during nighttime. To our best knowledge, this is the first study evaluating a distributed alarm system with not only patient monitor alarms but also alarms from ventilator, infusion pumps and patient call, giving a general overview of alarms at the handheld in an ICU environment. In literature, studies usually only evaluate patient monitor alarms, including less urgent alarms, using different alarm limits [e.g. 6,7]. Comparison of alarm pressure with literature is therefore not possible. However we can compare our two units, though having a different patient population, in order to improve the alarm chain. For alarm reduction, the ventilator at ICU should be connected via the patient monitor. For NICU, smarter patient monitor algorithms should be implemented [8]. Data mining techniques to investigate for example time coincidences of alarms would be useful as a basis for the development of smarter algorithms. In conclusion, an alarm management system can be used to filter the alarms and reduce the experienced alarm pressure. Further research is necessary to determine the clinical relevance of the various types of alarms [6] and to evaluate the experience of the medical and nursing staff. References 1. Borowski M., Gorges M., Fried R., Such O., Wrede C., Imhoff M.: Medical device alarms. Biomed Tech (Berl) 56, 73-83 (2011). 2. Blum J.M., Tremper KK. Alarms in the intensive care unit: too much of a good thing is dan-gerous: is it time to add some intelligence to alarms? Crit Care Med 38, 702-703 (2010). 3. Mitka M. Joint commission warns of alarm fatigue: multitude of alarms from monitoring devices problematic. JAMA 309, 2315-2316 (2013). 4. Keller JP,Jr. Clinical alarm hazards: a "top ten" health technology safety concern. J Electro-cardiol 45, 588-591 (2012). 5. Edworthy J, Hellier E. Alarms and human behaviour: implications for medical alarms. Br J Anaesth 97, 12-17 (2006). 6. Siebig S, Kuhls S, Imhoff M, Gather U, Scholmerich J, Wrede CE. Intensive care unit alarms--how many do we need? Crit Care Med 38, 451-456 (2010). 7. Ahlborn V, Bohnhorst B, Peter CS, Poets CF. False alarms in very low birthweight infants: comparison between three intensive care monitoring systems. Acta Paediatr 89, 571-576 (2000). 8. Vergales BD, Paget-Brown AO, Lee H, Guin LE, Smoot TJ, Rusin CG, et al. Accurate automated apnea analysis in preterm infants. Am J Perinatol 31, 157-162 (2014).