Continuous Surveillance of Sleep Apnea Patients in a Medical-Surgical Unit

Research Continuous Surveillance of Sleep Apnea Patients in a Medical-Surgical Unit

Dana Supe, Leah Baron, Tom Decker, Kyle Parker, Jeanne Venella, Sarah Williams, Kari Beaton, and John Zaleski

About the Authors Abstract Increasingly, surgical patients present with This report consists of two separate studies on the complex medical conditions and multiple Dana Supe, MD, MBA, DABSM, is use of continuous capnography monitoring comorbidities that make perioperative care program director conducted in an effort to improve patient safety more challenging. Patients often are pre- for clinical patient at Virtua Health System. The desire for improved scribed postoperative pain medications, safety at Virtua patient safety is motivating continuous monitor- mainly opioids, to control pain associated Health in Mount ing and improved surveillance in clinical areas with surgical procedures. These pain medica- Holly, NJ, and tions can produce undesirable and potentially medical director of operations at not traditionally equipped for such monitoring. Virtua Memorial Hospital of Burlington We explored the use of remote monitoring of life-threatening adverse effects. County. Email: [email protected] capnography, using enterprise middleware, in Virtua Health System (VHS) sought to patients recovering from surgery in a medical- prioritize narcotic safety by implementing Leah Baron, MD, surgical unit. Continuous monitoring noninvasive capnography monitoring in 2013.1 is chief of the department of traditionally has been used in higher-acuity As part of this narcotic safety program, an anesthesiology at settings, such as intensive care units. Patients approach for remotely monitoring and Virtua Memorial diagnosed or suspected to have obstructive or identifying respiratory depression was sought. Hospital in central sleep apnea may benefit from the Respiratory depression associated with use of Mount Holly, NJ, increased surveillance afforded by continuous opioids in the postoperative period is a and a member of the Burlington well-recognized adverse effect.2–4 Close Anesthesia Associates/Mednax. monitoring. Pain management in this cohort of Email: [email protected] patients, recovering from bariatric, joint replace- monitoring of these patients is suggested as a ment, or other major surgery, often involves means to identify early deterioration. Tom Decker is a administration of opioids (e.g., hydromorphone, Capnography is defined as the “noninva- systems integration morphine sulfate), which are known to increase sive measurement of the partial pressure of architect at Virtua Health in Marlton, risk of respiratory depression. Continuous carbon dioxide in exhaled breath,” and NJ. Email: tdecker@ monitoring of these patients increases the through its use, apnea can be detected almost virtua.org likelihood of detecting adverse clinical events. Our instantaneously.5 Recent studies identified goal was to implement continuous monitoring in capnography as providing a more sensitive and early predictor of opioid-induced respira- Kyle Parker is a order to identify alarm conditions caused by systems integrator adverse clinical events requiring intervention tory depression (OIRD), particularly in at Virtua Health (e.g., opioid-induced respiratory depression) and patients with obstructive sleep apnea (OSA) in Marlton, NJ. artifacts related to patient movement, suspect or central sleep apnea (CSA), and capnogra- Email: measurements, or other medical device–generated phy has been recommended as a best [email protected] alarm signals. practice for monitoring these adverse effects.6–11 Many postsurgical patients, who are at risk for OIRD, are cared for in medical-

236 Biomedical Instrumentation & Technology May/June 2017 © Copyright AAMI 2017. Single user license only. Copying, networking, and distribution prohibited. Features surgical units (MSUs), where capability for receiving narcotics with capnography … is the close monitoring is not as readily available as most reliable detector of hypoventilation” and in higher-acuity settings, such as intensive that concerns regarding the potentially high care units (ICUs). number of false-positive alarms in the When setting up increased levels of postoperative patient may be ameliorated by Jeanne Venella, surveillance in MSUs, care must be taken to developing “algorithms blending pulse DNP, MS, RN, CEN, CPEN, is differentiate actionable from nonactionable oximetry and capnography to yield greater chief nursing 18 alarm signals, so as to minimize false alarms benefit with fewer false-positive events.” officer at Bernoulli and the associated potential for alarm fatigue. In 2009, the Emergency Nurses Associa- Enterprise, Inc., in Alarm signal annunciation and the triggering tion identified certain “conclusions and Milford, CT. Email: of false alarm signals is deemed a major recommendations about the use of capnogra- [email protected] patient safety concern, with the number of phy for procedural sedation and analgesia in Sarah Williams, RT, nonactionable alarm signals yielding false- adults and children in the emergency is senior product positive alarm rates considerably exceeding department,” recommending “etCO2 is a manager at actionable alarm signal quantities. Estimates more sensitive indicator of respiratory Bernoulli Enterprise, have indicated that these false-positive, depression than [arterial oxygen saturation] Inc., in Milford, CT. Email: swilliams@ nonactionable, clinically insignificant alarms or clinician assessment.”17 bernoullihealth.com account for 85% to 99% of all alarm signals.12 In the process of implementing remote Some approaches to false alarm signal capnography monitoring, VHS engaged in Kari Beaton, RN, reduction involve basing alarm signaling both an initial pilot and a follow-on clinical is director of thresholds on derived parameters, such as study. The initial pilot focused on remote enterprise solutions at Bernoulli the Integrated Pulmonary Index, which is a monitoring of capnography alarm signals Enterprise, Inc., combinatorial index based on four param- issued through middleware to telemetry in Milford, CT. eters.11 Although we do not dispute the value technicians within the three hospitals com- Email: kbeaton@ of such combinatorial measures, the objec- prising the VHS enterprise. At the end of the bernoullihealth.com tive of the current work was to assess the initial pilot, remote capnography monitoring John Zaleski, PhD, relative benefit of sustained and combinato- was discontinued to analyze data and make CAP, CPHIMS, is rial alarms using clinical measures, including recommendations as to its utility and value. chief analytics officer at Bernoulli end-tidal carbon dioxide (etCO2), spontane- Following this period of analysis, a subse- ous respiratory rate (or respiratory frequency quent clinical trial was conducted in which Enterprise, Inc., [f ]), pulse rate (PR; measured using periph- recommendations resulting from the assess- in Milford, CT. R Email: jzaleski@ eral capillary oxygen saturation [SpO ] cuff), ment of the initial pilot were implemented. A 2 bernoullihealth.com and arterial oxygen saturation (measured summary of the results of both the initial pilot using SpO2 cuff), as they provide clearer and follow-on clinical study is provided, along clinical meaning compared with indices. with conclusions and lessons learned. Techniques for mitigating nonactionable alarms also are recognized as being of Initial Pilot Study supreme importance to reduce clinician Noninvasive capnography monitoring was workload and improve patient safety.13 done using Capnostream 20 Bedside Hospitals are concerned about alarm fatigue Monitors (Covidien, Needham, MA). because “it interferes with patient safety, and Creation and remote communication of both it exposes patients … to grave harm.”14 It also middleware-generated and medical device– has been asserted that “a majority of alarm- generated alarm signals were accomplished related adverse events result in brain injury using an enterprise middleware rules engine or death, carrying a median claim of nearly (Bernoulli One Analytics Software; Bernoulli $500,000.”15,16 Enterprise, Inc., Milford, CT). Upon initial The American Society of Anesthesiologists rollout of capnography monitoring, both in 2009 stated that “end-tidal carbon dioxide monitor- and middleware-generated alarm monitoring is more likely to detect hypercap- signals were communicated remotely to nia/hypercarbia and respiratory depression telemetry monitoring units, where telemetry than are clinical signs.”17 The Anesthesia technicians (“tele-techs”) could assist in the Patient Safety Foundation recommended that management and monitoring of these “monitoring the ventilation of patients patients and relay events to nursing units

Biomedical Instrumentation & Technology May/June 2017 237 © Copyright AAMI 2017. Single user license only. Copying, networking, and distribution prohibited. Features

when alarm signals were received. Data were the data were stored and processed in real communicated over the hospital information time and alarm signal messages were commu- technology network using TCP/IP protocols. nicated and displayed within a telemetry In-room alarm signals (both audible and bunker–based dashboard display on a dedi- visual) from the monitors were not impeded cated computer monitor. The analytics or changed in any way and were allowed to software also received admission-discharge- annunciate per normal bedside monitor transfer transactions from the enterprise operation. Middleware-generated alarm electronic health record system to facilitate signals were created based on the raw patient association to the monitors. measurements obtained from the monitors. The workflow for data collection and display Alarm signals were created based on end- were as depicted in Figure 1. Data received by user–defined thresholds, which, when the bridges were communicated wirelessly to breached, were communicated as visual and the analytics software, where the data were audible alarms to a dashboard view presented stored and processed against predefined alarm to the tele-techs. signal thresholds. Following evaluation As alarm signals were created and commu- against stored rules, threshold breaches then nicated, tele-techs would monitor the were communicated to a web-based dashboard dashboard view, which identified the status of within the tele-tech bunkers. all patients on capnography. The tele-techs Upon arrival in postanesthesia care units would communicate the alarm event to the (PACUs) from surgery, patients were nursing units associated with the patient for attached to the monitors via pulse oximetry whom the alarm event was issued, as well as cuff and nasal cannula. The monitor assem- print a copy of the alarm event for retention blies with bridges attached were mounted on within the telemetry bunker (a room dedi- roll stands and wheeled to the patient cated for telemetry monitoring within each of bedsides in PACUs. There, the nurse would the three hospitals). barcode the patient’s wrist bracelet, barcode The monitor communicates device-issued the label on the bridge, verify the coupling on discrete alarm signals in the form of binary the barcode liquid-crystal display (LCD) “on/off” messages: an alarm either is or is screen, and validate the coupling of the not issued. A summary of the available bridge to the patient by pressing a button on measurements and device-issued alarm the barcode scanner. All data collected from signals is provided in Table 1 in the data the monitor were then transmitted to the supplement (available online at http:// analytics software until the patient was aami-bit.org). These alarm signals provide no discharged from capnography monitoring. information relative to severity of specific The workflow for associating the patient with values. Rather, they serve to indicate that an the monitor is shown in Figure 1 in the alarm of predefined threshold, based on online supplement. monitor-set threshold values, was breached. A dedicated Datalogic Gryphon I GM4100- In all cases, monitors were configured to the HC wireless barcode scanner was used for same on-board threshold settings, and these the patient association process. This barcode values remained unchanged throughout the scanner has an LCD screen that allows for course of both the initial pilot and the the display of patient identifiers and the follow-on clinical study. In contrast, middle- bridge serial number. A workflow was ware-generated alarm signals were created to created whereby the identity of the patient, communicate threshold breaches that the assigned name of the bridge, and a contained both the threshold and the particu- confirmation request were displayed on the lar value causing the threshold breach. Gryphon LCD screen with which the nurse Data were captured from the monitors could interact. The purpose of this coupling using Bernoulli Serial-to-Ethernet Bridges was to assign patient name and identifiers to (Figure 1). The data collected through the the data collected from the monitor so that all bridges were communicated wirelessly to alarm signals and data would be issued Bernoulli analytics software, which was unambiguously and displayed through the installed in the enterprise data center. There, tele-tech dashboard. The dashboard view is

238 Biomedical Instrumentation & Technology May/June 2017 © Copyright AAMI 2017. Single user license only. Copying, networking, and distribution prohibited. Features

10× per minute: • etCO2 Capnostream monitor Bernoulli bridge • Pulse rate Hospital system • SpO2 • Respiration rate ADT • Device-issued alarms feed

Rules engine and alarm communication

Barcode used Upon event: for patient-to- • Threshold breaches device • Technical association alarms (initial pilot only) Data Dashboard display: storage alarms shown SpO cuff etCO2 nasal cannula 2 and graphically in grid rules view library

Figure 1. Capnostream 20p monitors communicated through Bernoulli IDM 3400 serial-to-ethernet bridges. Bridges would poll the monitor at the rate of of 10 times per minute and communicate data wirelessly to the Bernoulli Analytics Software located within the enterprise data center. Rules were maintained within the Bernoulli data storage and rules library and recalled for application against the received data. The serial-to-ethernet bridges were outfitted with barcodes to facilitate a patient-to-device association workflow using a Gryphon barcode reader. Also shown are the data collection and alarm signal communication from the monitors to the dashboard displayed within the telemetry bunkers. Abbreviations used: ADT,

admission discharge transfer; etCO2, end-tidal carbon dioxide; SpO2, peripheral capillary oxygen saturation.

shown in Figure 2 in the online supplement. measured from the monitor are shown, and

This view is created by the analytics software included etCO2, fR, SpO2, and PR. Measure- in the form of a web page that is displayed ments were initiated in PACUs. Both etCO2 within the telemetry bunkers for continuous and fR measurements were made using monitoring purposes. The green rectangle FilterLine etCO2 Sampling Lines (Medtronic, indicates an active patient for which no alarm Minneapolis, MN) designed for use with signals are being issued and contains Microstream-enabled capnography monitors, information pertaining to the patient, and measurements of PR and SpO2 were including name and identifiers. made using Nellcor finger-based pulse The clinical user to which the patient is oximetry sensors (Medtronic). assigned can also be optionally displayed. The rules and thresholds detailed in the Clicking on the green rectangle causes a vital next section were applied to the data received display box to be made visible. This vital from the monitors and visually depicted in display box depicts the current value of the telemetry dashboard view. Threshold measurements together with any issued breaches that corresponded to clinically alarms. Note that the findings in this display significant cautionary or low-priority levels can be tailored to the purposes of the clinical (i.e., breaches that corresponded to a caution- end user. In this case, those parameters ary level of concern) were shown in yellow

Biomedical Instrumentation & Technology May/June 2017 239 © Copyright AAMI 2017. Single user license only. Copying, networking, and distribution prohibited. Features

within the dashboard (Figure 3 in the online patient is eating). In such instances, a supplement). The specific text associated notification would be displayed on the with the middleware-generated alarm signal dashboard indicating that the patient was event is customizable and is contained within away from the monitor. Clinical staff elected the rule maintained by the analytics software. for such notifications to be displayed in Clicking within the vital display box purple (Figure 7 in the online supplement). triggered the flowsheet view (Figure 4 in the This was done to remove any confusion and online supplement). The data may be viewed to prevent any nonactionable alarm signals graphically or in tabular format. Data were associated with sensors removed from collected and displayed for the duration of patients being sent to the telemetry units. the patient encounter. Data that exceeded urgent (i.e., high-prior- Rules, Alarms, and Alarm Settings ity) levels based on the predefined rules Table 2 in the online supplement summa- associated with a specific measurement rizes the monitor alarm condition limit appear as a red rectangle in Figure 5 in the settings on key clinical parameters. These online supplement. Of note, the color scheme parameters remained unchanged from the adopted (green = normal; yellow = warning or initial pilot through the follow-on clinical low priority; red = urgent or high priority) was study. Note that while the alarm condition entirely customizable. The selection of the limit settings are most often referred to as threshold levels, color schemes, and audible high and low priority, the nomenclature used notifications were defined by the clinical team here (i.e., “urgent” and “caution”) was that prior to deployment and initiation of the study. preferred by the clinical staff. Note that in the case of the alarms displayed in Data were collected and processed at the this figure, a second notification is shown: that rate of 10 sets of measurements per minute. of an alarm threshold breach. This threshold The dashboard-displayed alarm signal events breach was defined using the analytics presented breaches of the clinically relevant software as an indicator when a value thresholds as defined in Table 2 in the online exceeded a specific threshold. Thus, this supplement. However, although cross-over threshold breach could be displayed, a custom was recognized between technical and alarm signal threshold breach associated with clinical alarm signals (i.e., alarm signals a rule could be displayed, or both could be intended for direct intervention by the displayed simultaneously. Again, clicking on clinical end user [nurse or respiratory the vital display box causes the flowsheet to be therapist]), technical alarm signals also could displayed (Figure 6 in the online supplement). cause and trigger clinical alarm signals. Also, This figure displays the occurrence of a low in certain cases (e.g., flow blockage, discon-

etCO2 threshold breach occurring at a meas- nects), the clinical end user must be made urement level of 15 mmHg. aware of these so as to intervene and correct. The alarms as defined using the rules The initial rollout involved an evaluation engine provide for both a color metaphor period of 2 weeks, after which the methods, indicating urgency (e.g., red color) and a workflow, and impressions of clinical and numeric value, both of which are customiz- technical staff were analyzed. A summary of able. For the purpose of this study, green the alarm and data findings are provided in levels were identified with an alarm level of 0, Table 3 in the online supplement. Not all indicating no alarm condition. Low-priority patients who were placed on capnography levels were associated with an alarm level of monitoring remained for the duration of 1, indicating cautionary or moderate urgency. their stays in MSUs, and in certain cases, Finally, high-priority levels were associated patient data were disqualified for technical or with an alarm level of 2, indicating immedi- clinical reasons (e.g., suspect measurements, ate urgency. technical glitches, patients removed from Technical alarm signals also were captured capnography for clinical reasons). Patients and communicated visually. For example, ranged widely in terms of duration on during the course of patient care, nursing capnography monitoring, from as little as an might need to pause the monitor (e.g., when hour to several days. The summary results

240 Biomedical Instrumentation & Technology May/June 2017 © Copyright AAMI 2017. Single user license only. Copying, networking, and distribution prohibited. Features shown in Table 3 in the online supplement high respiratory rate alarm signals domi- are for patients whose data were not disquali- nated. At the end of this evaluation period, it fied based on the criteria of suspect was determined that the quantities of alarm measurements, technical glitches, or prema- signals being issued to the telemetry units ture removal from capnography monitoring. were so high as to make the process of To illustrate the character of the findings, telemetry monitoring of these patients unten- time-based plots of the data associated with able using the method of communicating A decision was made one patient (patient 14) are shown in Figure 2 measurement threshold breaches. Hence, a to investigate ways in

(fR and etCO2) and Figures 8 and 9 in the decision was made to investigate ways in which to reduce alarm online supplement (PR and SpO ). The which to reduce alarm signals and provide 2 signals and provide only accompanying thresholds are displayed as only actionable notifications to appropriate overlays on these plots to add context to the clinical staff. actionable notifications raw measurements. A key point to be communicated in regard to appropriate clinical The expectation at the outset of the initial to monitoring patients receiving intravenous staff. pilot was that the alarm signals issued on or oral opioid administration is that a each patient would be manageable and substantial amount of artifact results from informative. This assumption was quickly patient movement. The nasal cannula and dismissed in practice, as is evident from the SpO2 sensors are a prime source of the summary shown in Table 3 in the online technical (and ultimately clinical) alarm supplement in which, in some cases, as signals. For example, movement of the nasal many as 427 alarms per hour were issued on cannula can result in the issuing of false a patient (patient 12) corresponding to readings, particularly if adjusted by the threshold breaches (principally) in low patient so that only one nostril is recorded. respiratory rate and low etCO2. An overall Patients report these cannula to be particu- average across all patients of 182 alarms per larly uncomfortable and often were required hour was determined from the raw counts of to wear them for many hours, leading to the data. By far, low etCO2 and both low and patients (or families) adjusting the tubing.

100

0 0 15 30 45 60 75 90 105 120 135 150 165 180 195 210 225 240 255 270

time after arrival from operating room (minutes)

Respiratory Rate (breaths/min) end-tidal CO2 (mmHg)

Figure 2. Combined respiratory rate (fR) and end-tidal carbon dioxide (etCO2) measurements for patient 14 in the initial pilot. Apnea events can be seen between 15 and 20 minutes, briefly at 38 minutes, briefly between 70 and 72 minutes, and for the extended period between 245 and 250 minutes, when both etCO2 and fR dropped to 0.

Biomedical Instrumentation & Technology May/June 2017 241 © Copyright AAMI 2017. Single user license only. Copying, networking, and distribution prohibited. Features

Thus, alarm signals associated with artifact case of sustained threshold breaches, a set of occurred. The sounding of alarms at the adjacent measurements will breach a thresh- point of care was intended to awaken or old and “trend” at or below the threshold for otherwise notify the patient and/or family of a particular period of time. They may situations that might be dangerous to the eventually self-correct or may not self-correct. patient (e.g., cessation of breathing), which While the individual threshold breaches can occur in cases OSA or CSA. However, as could be due to artifact, they also could be the was observed, some patients who were asleep harbinger of a true event. The decision to or sedated would not respond to these intervene when non–self-correcting measure- audible in-room alarm signals. ments appear is a clinical one, and the duration of the trend at or beyond the Postpilot Data Investigation threshold also is a question of clinical A poststudy investigation of the data was judgment. The exact cause of such threshold performed to gain insight into alarm signal breaches is not known unless an actual source and, ultimately, to assist in aiding observation of the patient takes place during reduction. The most immediate observation or surrounding the occurrence of the event to was that the vast majority of the alarm validate and verify cause. In the case of signals resulted from threshold breaches of individual measurements exceeding a two or more consecutive measurements. threshold, the cause could be bad measure- These consecutive alarm breaches, also ments, movement of the patient, or issues termed sustained alarm signals, could be used with the bedside monitor. In the case of The decision to to filter out certain types of artifact that were sustained measurements exceeding a intervene when spurious and nonrepeating. Of note, a single threshold, the cause could, again, be bad non–self-correcting sustained alarm might consist of multiple measurements, movement of the patient, a measurements appear instances of measurements breaching a true patient event, or issues with the bedside is a clinical one, and the predefined threshold. Thus, a single sus- monitor. tained alarm would be issued when multiple To validate the hypothesis that sustained duration of the trend at measurements breached the individual alarm signal generation would considerably or beyond the threshold thresholds of a single parameter. In the case reduce the overall number of alarm signals also is a question of of instantaneous alarm signals, these were issued, the data of the initial phase pilot were clinical judgment. counted as individual alarms when they were retrospectively evaluated against sustained distinctly issued, nonconsecutive events. delays of 30 seconds. The results are shown The clinical team hypothesized that in Table 4 in the online supplement. As individual, self-correcting measurements hypothesized, the overall number of events (i.e., those that breached a threshold, then per hour dropped to less than one-third those returned to normal range) should not be reported in Table 3 in the online supplement, communicated. Rather, only those instances with a maximum number of 122 alarm where measurements continuously trended signals per hour (patient 14). below/above a specified threshold for a However, the nagging clinical question predefined period of time should be commu- surrounding this mathematical reduction in nicated. This sustained alarm signal alarm signals was how many measurements communication is frequently used by should be sustained before reporting an physiologic monitors (and the subject event as clinically actionable to a care monitor) to reduce the likelihood of noise provider? A clinical discussion and survey of being communicated. Figures 3 and 4 the literature resulted in a decision to illustrate the concept of self-correcting consider 30 seconds as the sustained alarm

threshold breaches of etCO2 versus a sus- threshold. That is, if measurements in any

tained etCO2 threshold breach, respectively. individual parameter were sustained at or In the case of self-correcting threshold below/above the threshold value for 30 breaches, individual measurements may seconds or longer, then an alarm signal exceed an identified parameter threshold, but should be issued on an individual parameter. the following measurement will correct to In addition, however, alarm signals should the normal, within-threshold value. In the only be issued when the data are known to be

242 Biomedical Instrumentation & Technology May/June 2017 © Copyright AAMI 2017. Single user license only. Copying, networking, and distribution prohibited. Features

25 (mmHg)

20 etCO2

15 Threshold

0 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00 Time (minutes)

Nonconsecutive etCO2 measurements that exceeded threshold Figure 3. Example of self-correcting threshold breaches: individual, nonadjacent measurements exceeding a specified end-tidal carbon dioxide (etCO2) threshold. These occurrences may have been due to noise, bad measurements, or a combination of both.

25 (mmHg)

20 etCO2

15 Threshold

0 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00 Time (minutes)

Consecutive parameter measurements that exceed threshold Figure 4. Example of sustained threshold breach: Measurements do not self-correct to the normal range. This behavior may be indicative of an actionable event or may result from a systematic problem, such as suspect measurements.

Biomedical Instrumentation & Technology May/June 2017 243 © Copyright AAMI 2017. Single user license only. Copying, networking, and distribution prohibited. Features

valid. That is, if a known technical alarm signals, a series of rules was developed condition (e.g., nasal cannula off patient, requiring that multiple conditions be met pulse oximetry cuff off patient, calibration before middleware-generated alarm signals error) occurs, these technical alarm condi- were transmitted to clinical staff. These tions should be taken into account via rules were reviewed and tested on simulated communication to the clinical engineering data. The rules combined characteristics of staff and any data collected during a period in the measured parameters based on a study which such suspect measurements were of the data collected from the initial rollout obtained should not be used to calculate and from the literature. The follow-on study whether a sustained clinical alarm signal received approval from the VHS should be issued on a patient. Thus, several Institutional Review Board (reference questions were identified: identifier G15020), and informed consent • Is the selected duration of the sustained was obtained from all participants. alarm delay sufficient to reduce alarm signal traffic while not concomitantly Patient Eligibility introducing a patient safety concern? This clinical trial was designed to evaluate the • Which alarm signals should be communi- use of alarm signals generated using sus- cated to care providers? tained and combinatorial alarm rule • How long should alarm signals be commu- conditions over a period of 4 weeks at one nicated before escalating? hospital within the health system (Virtua • Can other measures or combinations of Memorial). Patients were enrolled in the study data provide an early indicator of patient based on existing diagnoses of OSA or respiratory compromise? meeting the STOP-BANG criteria for OSA (Table 5 in the online supplement). A total of A discussion regarding how best to validate 31 patients were recruited during this period, these questions led to the conclusion that four of whom were disqualified due to answering them in one investigation may not irregularities in data collection or clinical be possible. However, it may be possible to issues. In addition, one patient was discharged further quantify the findings to provide from the PACU and one patient transferred greater insight into the management of this from the PACU to the ICU. Thus, a total of 25 cohort of patients, which can then lead to patients were placed on capnography monitor- answers. Thus, a follow-on study was ing in the PACU. A summary of the patient formulated. This new pilot study would be population (17 women and 8 men, mean age focused on attempting to quantify the alarm 60 years) is provided in Table 1. Most patients signals associated with the selected sustained received intravenous hydromorphone or delays. To gain feedback on causality and morphine sulfate for pain management. For validity of the alarm signals, dedicated this follow-on study, the dashboard display nursing research staff would receive the was replaced with a direct communication to alarm signals via phone and would proceed research nursing using Voice-over-Internet- to each patient to observe and validate or Protocol (VoIP) phones (Cisco). otherwise identify the root cause of the alarm signal condition. Data Collection Workflow Data collection and processing workflow were Follow-on Clinical Study unchanged from the initial pilot. However, Recognizing the prevalence of artifact and communication workflow was changed the large quantity of alarm signals that can (shown in Figure 10 in the online supple- result, a decision was taken to perform a ment). The workflow was similar to that of the follow-on study with human participants. A initial pilot study, except that the processed steering committee of clinical and technical alarms were communicated in text from the staff was brought together to study and analytics software to the VoIP phones. When evaluate methods to reduce alarm signals an alarm signal was transmitted to the phone, and to communicate only those that were patient name, identifier, location, and cause of deemed truly actionable. To reduce alarm the alarm signal were displayed on the phone

244 Biomedical Instrumentation & Technology May/June 2017 © Copyright AAMI 2017. Single user license only. Copying, networking, and distribution prohibited. Features

LCD screen. After receiving the alarm signals, Age research nurses visited patients. Patients Patient Sex (years) Opioid Dosing requiring intervention were referred to the 1 Female 76 PCA hydromorphone floor nursing staff, or in the case of a need for 2 Female 42 Hydromorphone q2h rapid response, the standard intervention 4 Male 60 Hydromorphone q4h process was followed. 5 Female 76 PCA hydromorphone Sustained and Combinatorial Rules 6 Female 63 PCA hydromorphone and Alarm Signals Calculation 7 Female 47 Hydromorphone q3h The clinical team considered the use of 8 Male 61 PCA hydromorphone sustained (or persistent) and combinatorial (or 9 Female 41 Hydromorphone 1 mg q2h; multicriteria) alarms to identify actionable hydromorphone 2 mg q4h events. In the case of sustained alarms, criteria 10 Female 67 PCA hydromorphone for reporting necessitated that a condition be 11 Male 61 PCA hydromorphone maintained for a predetermined period of 12 Female 63 Hydromorphone q2h time. Clinical team members hypothesized 13 Male 62 PCA hydromorphone prior to the initiation of the clinical study that a sustained alarm signal delay of 30 seconds 14 Male 70 Hydromorphone q3h was a key measure of whether an alarm 15 Female 56 Hydromorphone q1h originating from monitors was actionable. 16 Female 72 PCA morphine Therefore, an alarm condition would not be 17 Male 58 Hydromorphone q2h signaled to research nursing staff unless it 18 Male 56 Hydromorphone q4h persisted for at least 30 seconds. 19 Female 83 PCA morphine Combinatorial alarms are defined as those 20 Female 70 Hydromorphone q4h for which multiple criteria must be met simultaneously before an alarm condition is 21 Male 49 PCA hydromorphone signaled. For example, a given parameter or 22 Female 29 Hydromorphone q2h set of parameters must meet specific condi- 23 Female 66 Hydromorphone q2h tions simultaneously for the combinatorial 24 Female 43 PCA hydromorphone alarm signal to be issued. Examples include 25 Female 57 Hydromorphone q2h no-breath events combined with low pulse Table 1. Characteristics of the study patient population. Abbreviation used: oxygen saturation events. PCA, patient-controlled analgesia. Figure 5 illustrates how combinatorial alarms can reduce alarm signal quantity. Plots of individual parameters (quantity is breaches could constitute a clinically signifi- arbitrary and is taken as the number N) are cant event. In this simplified illustration, six plotted against time. Each of these param- (shaded circles) of the overall 16 (shaded plus eters is displayed with threshold overlays on clear circles) threshold breaches occur each respective plot. A signal value exceeding simultaneously, thereby reducing the total a specified threshold constitutes a threshold quantity of issued alarm signal events from breach. Individually, the quantity of threshold 16 to six. breaches could translate into alarm signal All alarm signals (i.e., sustained individual notifications per each parameter, with the alarms, alarm signals calculated based on total quantity of issued alarm signals being combinatorial criteria) were communicated to the sum of all threshold breaches for each research nursing staff. As might be expected, parameter. Thus, a total of 16 threshold the total number of sustained alarm signals breaches are counted, corresponding to a outweighed combinatorial alarm signals by a total of 16 alarm signal events. The simulta- wide margin. The results of both types of neous occurrence of these threshold breaches alarm signals are reported below. can form the basis for a combinatorial alarm As patients were selected to participate in signal associated with all N parameters. the study on the basis of passing the STOP- Depending on the specific parameters, the BANG criteria, it was anticipated that some simultaneous occurrence of these threshold patients would experience a respiratory-

Biomedical Instrumentation & Technology May/June 2017 245 © Copyright AAMI 2017. Single user license only. Copying, networking, and distribution prohibited. Features

related adverse event. The literature suggests without hypoxia, defined here as the

that most incidents of hypoxic events are combinatorial assessment of etCO2 less 19 preceded by respiratory depression ; thus, a than 15 mmHg, fR fewer than 6 breaths/

decision was made to include several specific minute, and SpO2 less than 85%. combinatorial calculations in the assessment A modified form of bradypneic hypoventi- to evaluate how our data compared with lation (mBHH) alarm signal, both with identified combinatorial alarm conditions and without hypoxia, defined here as the

and alarm conditions reported in the litera- combinatorial calculation of etCO2 greater

ture. These included: than 65 mmHg, fR fewer than 6 breaths/

1. Calculated apnea (30 seconds) alarm signal minute, and SpO2 less than 85%.

based on measured etCO2 and fR. 3. Sustained alarms on etCO2, pulse rate,

2. Calculated no-breath and no-pulse combi- SpO2, and fR, whereby alarms would only natorial alarm signal. be issued in the event of a non–self-cor- A modified form of hypopneic hypoventila- recting threshold breach of 30 seconds’ tion (mHHH) alarm signal, both with and minimum duration.

Of note, bradypneic and hypopneic Parameter 1 versus time hypoventilation are sometimes referred to as 60 type 1 and type 2 hypoventilation, respec-

50 tively, and, as the sensitivity of etCO2 has 40 been documented to precede the onset of a 30 hypoxic event, these are clinically significant 11,19 20 findings that may require intervention. 10 In all cases, technical alarm signals were 0 filtered or otherwise removed from the 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80 2.00 2.20 2.40 2.60 2.80 3.00 3.20 3.40 3.60 3.80 4.00 4.20 4.40 4.60 4.80 5.00 clinical alarm signal reporting. However, a report of all alarm signals issued from the monitors is provided in the following section. Parameter 2 versus time 40 Findings Continuously measured parameters included 30 arterial oxygen saturation measurement via 20 finger-based SpO2, etCO2, PR, and fR. Key 10 metrics included sustained levels of hypoxia (SpO2 ≤85%), hypocarbia (etCO2 ≤15 mmHg) 0 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80 2.00 2.20 2.40 2.60 2.80 3.00 3.20 3.40 3.60 3.80 4.00 4.20 4.40 4.60 4.80 5.00 and hypercarbia (etCO2 ≥65 mmHg), respira-

tory rate (fR ≤6 breaths/minute; fR ≥24 breaths/ minute), and bradycardia (pulse rate ≤40 bpm) Parameter N versus time and tachycardia (pulse rate ≥150 bpm). Data collected from the monitors showed 80 that low etCO2 thresholds were breached in 60 5% of measurements, while fewer than 1% of measurements exceeded the high etCO 40 2 threshold of 65 mmHg. Approximately 6% of 20 measurements fell below the low fR threshold of 6 breaths/minute, and 15% fell below 8 0 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80 2.00 2.20 2.40 2.60 2.80 3.00 3.20 3.40 3.60 3.80 4.00 4.20 4.40 4.60 4.80 5.00 breaths/minute. Fewer than 1% of measure-

ments fell below SpO2 threshold of 85%. Of Figure 5. Multiple measurements plotted against time. Combinatorial alarm signals are issued all PR measurements taken, fewer than 1% only when the threshold criteria for all parameters are met simultaneously. The combinatorial fell below the bradycardia threshold of 40 alarm signal can have the effect of reducing the overall alarm reporting frequency due to false bpm. No patients experienced high PR alarm alarms because the potential for individual measures all meeting their threshold criteria simultaneously is hypothesized to be more likely associated with a true-positive event than false limit breaches. alarm signals. The correlation coefficient between

246 Biomedical Instrumentation & Technology May/June 2017 © Copyright AAMI 2017. Single user license only. Copying, networking, and distribution prohibited. Features

machine-issued low etCO2 alarm signal Parameter Type No. counts and low f alarm signal counts was R Respiratory rate determined to be 0.95. That is, low etCO 2 ≤6 breaths/minute 7,947 correlated highly with low fR, which might be expected in a patient population predisposed ≥28 breaths/minute 6,750 to OSA. No significant correlation was found SpO2 ≤85% 880 between low SpO2 and low etCO2, including monitor-issued no-breath alarms (i.e., fR = 0), Pulse rate and it is hypothesized that because these ≤40 bpm 10 patients were monitored so closely and SpO 2 ≥150 bpm 0 drops relatively slowly compared with etCO , 2 etCO interventions occurred before hypoxia was 2 experienced by most patients within the ≤15 mmHg 7,221 study population. ≥65 mmHg 4 A total of 193,177 data points were measured Total data points per 193,177 parameter per parameter. Average fR was 15 ± 6 breaths or respirations per minute, average etCO was 36 2 Table 2. Distribution of measurements exceeding monitor alarm settings. All technical alarms ± 9 mmHg, average pulse rate was 72 ± 15 were removed. Threshold breaches were dominated by respiratory rate and low end-tidal carbon dioxide (etCO ). Abbreviation used: SpO , peripheral capillary oxygen saturation. bpm, and average SpO2 97 ± 3%. Key findings 2 2 of middleware-generated alarm signals are summarized in Tables 2 through 4. Low Respiratory Rate SpO Pulse Rate respiratory rate presented the largest source of 2 etCO (mmHg) (breaths/minute) (%) (bpm) 2 alarm signals generated, followed by low Sustained etCO2. This tended to make sense given that Alarm the patient selection criteria weighted patients Signal heavily toward diagnosis of OSA (i.e., most of Delay the recruited patients had OSA). (seconds) ≤6 ≥28 ≤85 ≤40 ≥150 ≤15 ≥65 A parametric was run on the data to assess 18 6,739 3,888 257 6 0 5,361 2 the effect of sustained alarm delay (Table 3). 30 5,832 2,698 103 4 0 4,635 0 Measurements were collected every 6 42 5,242 2,070 54 2 0 4,189 0 seconds. Parametrically, as sustained delay was varied from 18 to 60 seconds, the 48 482 864 39 2 0 342 0 number of sustained alarm signals 60 405 705 20 0 0 284 0 decreased. Hence, as the data collection inter- Table 3. Middleware-generated alarm signals based on sustained alarm delays, parametrized val was a set of measurements every 6 against five different alarm delay levels. A sharp reduction in quantities of middleware- seconds, this meant that data collection generated alarm signals is achieved when increasing sustained delay from 42 to 48 seconds. Note that even with sustained alarm delay upwards of 60 seconds, the number of alarm occurred at 0, 6, 12, 18, 24, 30, 36, 42, 48, 54, threshold breaches remains considerably high (in excess of 1,000 total). Abbreviations used: and 60 seconds (corresponding to 0 seconds etCO2, end-tidal carbon dioxide; SpO2, peripheral capillary oxygen saturation. in the next minute). The drop in number of sustained alarm signals achieved between 42 Sustained Alarm Hypopneic Hypopneic and 48 seconds of sustained delay was Signal Delay Hypoventilation with Hypoventilation† significant (by one order of magnitude). (seconds) Hypoxia* (no.) (no.) Sustained alarm signals for single param- 6 106 4,852 eters remained high, even through 42 12 59 4,522 seconds of delay. The question arose as to 18 0 209 whether patient safety is put at risk by 24 0 0 increasing the delay time up to 1 minute.

During this clinical trial, clinical staff were Table 4. Combinatorial middleware-generated alarm signals based on respiratory rate (fR), peripheral capillary oxygen saturation (SpO ), and end-tidal carbon dioxide (etCO ). most comfortable with 30-second sustained 2 2 Combinatorial alarm signals (generated by calculating hypopneic hypoventilation both with delays and were not desirous of increasing and without hypoxia) were weighted primarily by low fR and low etCO2. Combinatorial alarm beyond that threshold. At 30 seconds of signals also were parametrized against sustained delays of 6, 12, 18, and 24 seconds. *fR ≤6 breaths/minute, SpO ≤85%, etCO ≤15 mmHg. †f ≤6 breaths/minute, etCO ≤15 mmHg. sustained delay, however, a large quantity of 2 2 R 2 alarm signals occurred that research nursing

reported to be mostly nonactionable. (Figure 11 in the online supplement). The The combinatorial rules produced alarm average respiratory rate determined in this signal quantities that were at least an order of cohort corroborated similar findings.20 magnitude and several factors fewer than the • A total of 5% of measurements fell below

sustained alarm signals on individual the 15 mmHg etCO2 threshold, while fewer parameters (Table 4). More significantly, the than 1% of measurements exceeded the combinatorial alarm signals also were hypercarbia threshold of 65 mmHg (Figure parametrized as sustained delays from 6 12 in the online supplement), indicating through 24 seconds. A 6-second combinato- that most of the thresholds breaches on

rial alarm implies two adjacent etCO2 were related to hypocarbia. This measurements that meet the threshold made sense to the researchers, as most criteria specified by the combinatorial alarm patients were diagnosed with OSA. rule. Of note, the quantities of combinatorial Further, low respiratory rate was the source alarm signals are significantly lower than the of most respiratory alarm signals (Figure single-parameter sustained alarm signals. 13 in the online supplement). Also, the total number of alarm signal • Pulse oximetry and pulse rate alarm signals quantities decreased much more quickly with occurred far less often than either respira-

sustained delay compared with the single- tory rate or etCO2 alarms (Figures 14 and parameter sustained alarm signals, as the 15, respectively, in the online supplement). criterion that multiple parameters must • No patients were found to have met the simultaneously meet the specified thresholds mBHH combinatorial alarm threshold, is much more stringent. though a number of patients met the mHHH threshold criteria (Table 4). Strong correlation was observed between middle- Patients were experiencing extended periods of single- ware-calculated mHHH alarm signals and parameter threshold breaches that were continuous or machine-issued low etCO2 alarm signals repetitive in nature. In most cases, these single-parameter alarm (mHHH and machine-issued low etCO2: 0.71; mHHH and machine-issued low f : signals did not signify clinically meaningful events requiring R 0.68; machine-issued low f and machine- intervention, as verified by research nursing staff. R issued low etCO2: 0.95). • Combinatorial alarm signals associated The results yielded the following with these events for 6- and 12-second observations: sustained delay are 106 and 59, respectively

• Sustained middleware-generated alarm (Table 4). Of note, when SpO2 was removed signals (i.e., persisting for 30 seconds or from the combinatorial calculations, the longer) remained quite high in this cohort, number of alarm signals was large (>4,500) even when the duration of the persistent until sustained delay of 18 seconds was delay was increased from 18 through 42 used, at which point the quantity of alarm seconds, with a steep reduction in alarm signals decreased to 209. In both of these signals achieved at 48 seconds of sustained combinatorial rule calculation cases, the delay (Table 3). A significant reduction in quantity of alarm signals was much more middleware-generated alarm signals manageable and provided a hint as to how occurred when the sustained delay was to improve alarm management. However, increased to 48 seconds. Patients were by using hypopneic hypoventilation experiencing extended periods of single- combinatorial alarm signals, more than a parameter threshold breaches that were 98% reduction over 30-second sustained

continuous or repetitive in nature. In most middleware-generated respiratory fR and

cases, these single-parameter alarm signals etCO2 alarm signals was achieved. did not signify clinically meaningful events • Finally, among seven patients who were requiring intervention, as verified by identified as requiring some form of research nursing staff. intervention, four had true respiratory • Low mean respiratory rate correlated with distress, with one patient requiring inter-

both low and high etCO2 measurements vention (i.e., administration of naloxone

hydrochloride to reverse the effect of the The distribution of all machine-issued opioid and placement on noninvasive alarm signals, considering both clinical and ventilator support). These patients were technical alarms, also is of interest. Table 5 discovered as a result of the sustained summarizes the machine-issued alarm alarms, and combinatorial alarms were signals issued cumulatively across the patient triggered for these patients as well. population. A significant number of nasal cannula and pulse oximetry sensor discon- Table 6 in the online supplement summa- nects were issued during the study. These are rizes the alarm signals issued by the monitors. clinically actionable. The large quantities A review of overall quantities of machine- resulted from the fact that the alarm was issued alarms revealed the following: communicated each time the monitor was • There were a total of 8,181 machine-issued polled until the condition was corrected. The

no-breath (7,065) and low-fR (1,116) alarm quantities, together with the type of technical signals compared with 5,832 middleware- alarm signals issued, motivate the need for a

generated low-fR and no-breath alarm workflow that integrates clinical engineering signals. The comparative middleware- into the alarm reporting infrastructure, so issued alarm signals were based on a that technical alarms (e.g., low battery 30-second sustained delay. Thus, to a notifications) can be communicated to staff degree, differences in reports can be to ensure that technical intervention takes attributed to differences in calculations and place in a timely manner. The existence of algorithms contained within the monitor calibration errors also suggests the need for a versus the rules created within the analytics software. By decreasing the sustained Alarm Condition No. delay to 18 seconds, the result is 6,739 ALR-DISC-SPO2 8,527 middleware-generated low-f alarm signals. R ALR-LO-BAT 8,301 • Machine-issued high-fR alarm signals totaled 2,173 compared with 2,698 middle- ALR-NO-BREATH 7,065 ware-issued alarm signals (based on ALR-LO-IPI 5,153 30-second sustained delay). ALR-LO-CO2EX 3,065

• Machine-issued low etCO2 alarm signals ALR-FL-DISC-CO2 3,051 totaled 3,065 compared 4,635 middleware- ALR-PR-NF 1,925 issued alarm signals (based on 30-second ALR-HI-RR 2,173 sustained delay). Again, by decreasing ALR-CO2-PUMP-OFF 1,768 sustained delay to 18 seconds, the result is 5,361 middleware-generated alarm signals. ALR-LO-PR 1,637 ALR-OFF-SPO2 1,318 • The quantities of SpO2 disconnects (8,527) and low battery (8,301) were significantly ALR-LO-RR 1,116 high, thereby suggesting the need for ALR-LO-SPO2 873 workflow improvements in terms of ALR-CO2-CHK-CAL 331 compliance and preparation of monitoring FL-BLOCK 308 devices before use. ALR-STBY-CO2 264 • CO flow disconnect alarm signals (3,051) 2 ALR-CO2-CHK-FLW 52 corresponded to the number of low etCO2 alarm signal reports. CO2-MLFNC 45 • Although the number of middleware- SPO2-MLFNC 37

issued etCO2 alarm signals exceeded that of ALR-HI-PR 9 machine-issued alarm signals, the alarm ALR-HI-CO2EX 3 signals based on combinatorial rules (Table ALR-HI-SPO2 0 4) suggested that the combinatorial ALR-STBY-SPO2 0 middleware-generated alarm signals can be Total 47,021 reduced greatly compared with either single-parameter sustained or machine- Table 5. Hierarchy of device-issued alarm signals, sorted from highest to lowest quantities issued alarm signals.

systematic quality control equipment turna- respiratory depression. Hence, the remotely round process by identifying the type and communicated alarms to the nursing phones number of monitors that fail to meet specifi- had the added effect of a safety net. cations before use on the next patient. Finally, to mitigate the effects of repetitive trended alarm signal fatigue, as was experi- Conclusion enced, the authors recommend using Two successive investigative studies were annunciated alarms based on combinatorial conducted to identify the utility and practical rules, particularly with this patient cohort, implementation of remotely communicated and to implement a protocol whereby sustained and combinatorial capnography response and escalation occurred promptly alarm signals, in order to reduce alarm signal upon receiving combinatorial alarm signals.

load and improve patient surveillance. The If etCO2 monitoring is to be used, then the results suggested that combinatorial alarm authors recommend thoroughly training clini- signals based on multiparameter assessment cal and biomedical staff, patients, and families reduced overall load better than individual- on its application, use, and limitations. n parameter sustained alarm signals and appeared to be more effective at identifying Acknowledgments at-risk patients. Refinement of these combi- To Virtua Health System for supporting the natorial alarm signals requires further study and to the following Virtua employees investigation to validate whether they can for their assistance: Anne Miller, RN, research The results suggested serve as a safe and effective means of detect- nurse; Jennylin Schott, RN, research nurse; that combinatorial ing respiratory depression among patients at Tom Decker, systems integration architect, alarm signals based risk for opioid-induced respiratory depres- medical device integration; and Kyle Parker, on multiparameter sion. It is encouraging that none of the systems integrator, medical device integration. monitored patients required rapid response assessment reduced team activation for events that were unde- Conflict of Interest overall load better than tected by either the sustained or Beaton, Venella, Williams, and Zaleski are individual-parameter combinatorial alarms. However, the authors employed by Bernoulli Enterprise, Inc., which sustained alarm signals acknowledge that the study population was was the middleware vendor for data collection and alarm processing used in the study. and appeared to be more limited and that further investigation in a larger population is merited. effective at identifying The number of patients needed to draw Financial Disclosure at-risk patients. valid conclusions is a matter of frequent Research nursing time associated with the debate. Pilot studies of comparable size have study was funded by Bernoulli Enterprise, been performed, thereby establishing some Inc. The authors did not receive compensa- precedent as to the validity of results concern- tion for research participation, and the ing this patient population.6,21,22 Further, the research was not grant supported. current population was highly targeted toward patients who were either diagnosed to have References OSA or who may have been susceptible to 1. Baron L, Decker T, Supe D, et al. AAMI respiratory depression (as assessed using the Foundation Safety Innovations: Virtua STOP-BANG criteria). This places the Health: Implementing Capnography in population bias more on the likelihood of Low Acuity Settings. Available at: http:// expecting, versus not expecting, apneic s3.amazonaws.com/rdcms-aami/files/ events. Considering that the objective of this production/public/FileDownloads/Foundation/ pilot study was to inform and refine assump- SafetyInnovation/2016_SI_Virtua_Capnography. tions, these results should help inform future pdf. Accessed March 27, 2017. rollout efforts. 2. Lenz G, Heipertz W, Epple E. Capnometry In-room alarm settings and audible alarms for Continuous Postoperative Monitoring of were not changed on any monitor. Somewhat Nonintubated, Spontaneously Breathing Patients. anecdotally, in a subset of patients, the J Clin Monit. 1991;7(3):245–8. in-room capnography alarms did not have the effect of arousing them from a state of

3. Stoelting RK, Weinger MB. Special Editorial: 13. AAMI Foundation. Clinical Alarm Management Dangers of Postoperative Opioids: Is There a Compendium. Available at: www.aami.org/ Cure? Available at: www.apsf.org/newsletters/ alarmcompendium. Accessed March 26, 2017. html/2009/summer/01_opiods.htm. Accessed 14. MacDonald I. Hospitals rank alarm fatigue as March 27, 2017. top patient safety concern. Available at: www. 4. Weinger MB, Lee LA. No Patient Shall Be Harmed fiercehealthcare.com/healthcare/hospitals-rank- By Opioid-Induced Respiratory Depression. alarm-fatigue-as-top-patient-safety-concern. Available at: www.apsf.org/newsletters/html/2011/ Accessed March 21, 2017. fall/01_opioid.htm. Accessed March 21, 2017. 15. Bennett CL, McNeer RR. PT-SAFE: A Software 5. Krauss B, Hess DR. Capnography for Procedural Tool for Development and Annunciation Sedation and Analgesia in the Emergency of Medical Audible Alarms. Anesth Analg. Department. Ann Emerg Med. 2007;50(2):172–81. 2012;114(3):576–83.

6. Soto RG, Fu ES, Vila H Jr, Miguel RV. 16. Olympio MA. APSF Workshop Recommends New Capnography Accurately Detects Apnea During Standards. Available at: www.apsf.org/newsletters/ Monitored Anesthesia Care. Anesth Analg. html/2004/winter/01workshop.htm. Accessed 2004;99(2):379–82. March 29, 2017.

7. Stocki D, Matot I, Einav S, et al. A Randomized 17. Oridion. Standards Mandating Capnography Controlled Trial of the Efficacy and Respiratory Monitoring. Available at: www.multileadmedics. Effects of Patient-Controlled Intravenous com/www.multileadmedics.com/Capnography_ Remifentanil Analgesia and Patient-Controlled files/capnogrpahy%20standards-final.pdf. Epidural Analgesia in Laboring Women. Anesth Accessed March 26, 2017. Analg. 2014;118(3):589–97. 18. Kodali BS. Capnography Outside the Operating 8. Overdyk FJ, Carter R, Maddox RR, et al. Rooms. Anesthesiology. 2013;118(1):192–201. Continuous Oximetry/Capnometry Monitoring 19. Deitch K, Rowden A, Damiron K, et al. Reveals Frequent Desaturation and Bradypnea Unrecognized Hypoxia and Respiratory During Patient-Controlled Analgesia. Anesth Depression in Emergency Department Patients Analg. 2007;105(2):412–8. Sedated For Psychomotor Agitation: Pilot Study. 9. Wong M, Mabuyi A, Gonzalez B. First National West J Emerg Med. 2014;15(4):430–7. Survey of Patient-Controlled Analgesia Practices. 20. McGrath SP, Pyke J, Taenzer AH. Assessment of Available at: www.premiersafetyinstitute.org/wp- continuous acoustic respiratory rate monitoring content/uploads/PPAHS-national-survey-patient- as an addition to a pulse oximetry-based patient controlled-analgesia.pdf. Accessed March 27, 2017. surveillance system. J Clin Monit Comput. 2016 10. Oridion. Monitoring End Tidal CO2 in Post May 3. [Epub ahead of print]. Surgical Patients: Continuous Breath Monitoring 21. Kyriacou PA, Jones DP, Langford RM, Petros AJ. A Increases Patient Safety. Available at: https:// Pilot Study of Neonatal and Pediatric Esophageal spotidoc.com/doc/279085/monitoring-end-tidal- Pulse Oximetry. Anesth Analg. 2008;107(3):905–8. co2-in-post-surgical-patients. Accessed March 26, 2017. 22. Restrepo RD, Otaibi AA, Wittnebel L, et al. Correlation between ABG Parameters and the 11. Garah J, Adiv OE, Rosen I, Shaoul R. The value Integrated Pulmonary Index in a Medical-Surgical of Integrated Pulmonary Index (IPI) monitoring ICU in Saudi Arabia. Society for Technology in during endoscopies in children. J Clin Monit Anesthesia Annual Meeting, Phoenix, AZ, 2013. Comput. 2015;29(6):773–8.

12. Association for the Advancement of Medical Instrumentation. Clinical Alarms 2011 Summit: Convened by AAMI, FDA, TJC, ACCE, and ECRI Institute. Available at: http://s3.amazonaws. com/rdcms-aami/files/production/public/ FileDownloads/Summits/2011_Alarms_Summit_ publication.pdf. Accessed March 21, 2017.

Biomedical Instrumentation & Technology May/June 2017 251