J Clin Monit Comput DOI 10.1007/s10877-014-9556-8
ORIGINAL RESEARCH
Detection of respiratory compromise by acoustic monitoring, capnography, and brain function monitoring during monitored anesthesia care Pedro P. Tanaka • Maria Tanaka • David R. Drover
Received: 30 June 2013 / Accepted: 7 January 2014 Ó Springer Science+Business Media New York 2014
Abstract Episodes of apnea in sedated patients represent a risk of respiratory compromise. We hypothesized that acoustic monitoring would be equivalent to capnography for detection of respiratory pauses, with fewer false alarms. In addition, we hypothesized that the patient state index (PSI) would be correlated with the frequency of respiratory pauses and therefore could provide information about the risk of apnea during sedation. Patients undergoing sedation for surgical procedures were monitored for respiration rate using acoustic monitoring and capnography and for depth of sedation using the PSI. A clinician blinded to the acoustic and sedation monitor observed the capnograph and patient to assess sedation and episodes of apnea. Another clinician retrospectively reviewed the capnography and acoustic waveform and sound files to identify true positive and false positive respiratory pauses by each method (reference method). Sensitivity, specificity, and likelihood ratio for detection of respiratory pause was calculated for acoustic monitoring and capnography. The correlation of PSI with respiratory pause events was determined. For the 51 respiratory pauses validated by retrospective analysis, the sensitivity, specificity, and likelihood ratio positive for detection were 16, 96 %, and 3.5 for clinician observation; 88, 7 %, and 1.0 for capnography; and 55, 87 %, and 4.1 for acoustic monitoring. There was no correlation between PSI and respiratory pause events. Acoustic monitoring had the highest likelihood ratio positive for detection of respiratory pause events compared with capnography and clinician observation and,
P. P. Tanaka (&) M. Tanaka D. R. Drover Department of Anesthesiology, Perioperative and Pain Medicine, Stanford University School of Medicine, 300 Pasteur Dr. H3577, Stanford, CA 94305-5640, USA e-mail: [email protected]
therefore, may provide the best method for respiration rate monitoring during these procedures. Keywords Apnea Capnography Monitoring: intraoperative Sedation
1 Introduction The number of procedures being performed under sedation, and commonly under very deep sedation, is increasing [1]. In many situations, administering sedative drugs and monitoring the patient is not done by an anesthesiologist, but by less trained clinical personnel. Even when an anesthesiologist performs deep sedation, there are a variety of reasons why he or she may be remote from the airway, making it difficult to assess the level of sedation and respiratory effort. Patient monitoring that allows for the timely detection of hypoventilation is necessary during these procedures. The American Society of Anesthesiologists has amended its Standards for Basic Anesthetic Monitoring to include mandatory exhaled end-tidal carbon dioxide (ETCO2) monitoring during both moderate and deep sedation [2]. Clinician observation, pulse oximetry, and capnography are used individually or in combination to monitor ventilation during sedation, but each method has known limitations. Clinician observation is unreliable owing to competing priorities during procedures [3], pulse oximetry may be a late indicator of hypoventilation especially when supplemental oxygen is administered [4], and capnography can have a high rate of false alarms [5]. Rainbow Acoustic MonitoringTM by Pulse CO-Oximetry is a relatively new method of monitoring ventilation, and has been shown to provide accurate estimations of respiration rate in the post
123
J Clin Monit Comput
anesthesia care unit [6, 7] and during procedural sedation for endoscopy [8]. The purpose of this study was to compare the accuracy of acoustic monitoring and capnography for detection of respiratory pauses during sedation as compared with a retrospective analysis of waveform and sound files analyzed by a clinician. As a secondary endpoint we examined whether apnea is associated with the level of sedation as determined by patient state index (PSI) during monitored anesthesia care (MAC). We hypothesized that during deep sedation it would be more likely for the airway to become obstructed and so tested whether the depth of sedation was associated with respiratory pause events.
2 Methods After receiving approval by the Stanford University Institutional Review board and written informed consent of patients, we conducted a prospective, single-center study of patients ASA I to III, 18 years of age or older, who were scheduled to undergo sedation with local or regional anesthesia. To facilitate insertion of the local or regional block, all patients received premedication consisting of midazolam and fentanyl administered intravenously (IV) at the discretion of the anesthesiologist or regional anesthesia team in the preoperative area. All patients received standard monitoring consisting of oxygen saturation by pulse oximetry (SpO2), noninvasive blood pressure measurement, end-tidal carbon dioxide monitoring (capnography, ETCO2) using tubing with a standard nasal cannula, skin temperature, and electroencephalogram recordings. An additional capnography monitor (CapnostreamÒ 20, Oridion Capnography, Inc., Bedford, MA, USA) was connected to the same nasal cannula to collect study data. Oxygen was administered continuously through a facial mask at a rate of 4–10 L/min. During the procedure, repositioning of the patient’s head, insertion of a nasal or oral airway and decreasing or increasing sedation was performed at the discretion of the anesthesiologist and based on the standard of care protocol. An adhesive acoustic respiration sensor (RAS-125TM revision C) was placed on the patient’s necks and connected to a Pulse COOximeter with Rainbow Acoustic Monitoring (Rad-87, software v. 7805, Masimo Corp., Irvine CA, USA) to provide acoustic respiration rate (RRa), SpO2, pulse rate (PR), and perfusion index (PI). A stethoscope was attached to the patient’s neck on the side contralateral to the adhesive acoustic sensor. A sensor array connected to a brain function monitor (SEDLine, Masimo Corp, Irvine CA, USA) was applied to the forehead of all patients to monitor the level of sedation, as indicated by the PSI. The capnograph, brain function monitor, and acoustic monitor data
123
were all electronically synchronized and connected to a laptop computer running data collection software (automatic data collection, Masimo Corporation) for continuous data recording. In patients who presented with a flat carbon dioxide line on the standard of care capnograph for 10 s, absence of airflow was confirmed by listening to the neck with the stethoscope by a clinician blinded to the acoustic monitor, study capnograph and SEDLine monitor. These events were recorded as respiratory pauses. A second clinician not blinded to the study monitors marked in the electronic record any respiratory pause event detected by each method. For the purposes of the study, a respiratory pause was defined as a cessation of breathing for C20 s. After the procedure, a clinician replayed the waveform and sound files recorded from the acoustic monitor and capnograph using a specialized LabView-based software program (Tag Editor, Masimo) to determine the validity of the respiratory pauses detected by each method. Tag Editor allows for the simultaneous viewing of both the acoustic and capnography waveforms while listening to the breathing sounds from the acoustic signal, to determine inspiration and expiration reference markers within the respiration cycle independently from the acoustic monitor or capnometercalculated respiration rates. This method has been used in another study investigating the accuracy of acoustic monitoring for detection of respiratory pauses [6]. Periods of approximately 1 min before to 1 min after each presumed respiratory pause were reviewed. Therefore, any respiratory pause event missed by all methods was not included in the analysis. Periods of respiratory pause were determined by (1) clinical judgment, (2) respiratory rate as determined by EtCO22 (RR ETCO2), and (3) RRa. If a patient required intubation based on standard of care, data collection was stopped for that patient but the data collected up to that point contributed to the data set. 2.1 Data analysis A power analysis was not feasible because there were no preliminary data on which to base the analysis. Thus, we deemed this to be a pilot study. To evaluate the agreement between capnography and acoustic monitoring for estimation of respiration rate, bias (mean of differences between methods), standard deviation of the bias, and limit of agreement were calculated and graphed using the Bland–Altman method [9]. To determine the accuracy of detection of respiratory pauses, we calculated the number of respiratory pauses detected that were validated (true positives), the number of respiratory pauses detected and not validated (false positives), and the number of true positive detected by each test method divided by the total number of true events.
J Clin Monit Comput
To determine the sensitivity and specificity for each method, we used the following definitions: A true positive was a presumed respiratory pause detected by a test method that was verified by retrospective analysis. A true negative was no respiratory pause detected by the test method or the reference method at a time when one or more other test methods detected a respiratory pause. A false positive was a presumed respiratory pause detected by a test method that was not validated by retrospective analysis. A false negative was a respiratory pause verified by retrospective analysis that was not detected by the test method. Sensitivity, therefore, was equal to the number of true positives divided by the number of true positives plus the number of false negatives. Specificity was equal to the number of true negatives divided by the number of true negatives plus the number of false positives. Confidence intervals (CI) for sensitivity and specificity for each test method were calculated by a bootstrapping technique, where a resampling with replacement was used with 1,000 samples from 118 available data points to provide an empirically derived distribution to estimate the upper and lower boundaries. Likelihood ratios were used to provide a single metric that combines sensitivity and specificity to compare the test methods. For the purposes of this study, a likelihood ratio positive (LR?) indicates the probability that the detected respiratory pause was a true event, with a larger number indicating a higher probability that a respiratory pause would be detected by the method. In contrast, a likelihood ratio negative (LR-) is the probability of a true respiratory pause not being detected by the test method, with the smaller number indicating a lower probability of a missed event. Likelihood ratios with upper and lower boundaries were calculated for each test method. To determine if there was a relationship between respiration rate and depth of sedation, a regression plot of RRa values to PSI values was plotted and the correlation of determination (R2) was calculated. Additionally, the distribution of PSI values verses the frequency of respiratory pauses was plotted.
A total of 137,199 paired ETCO2 and RRa respiration rate values were collected. The bias, standard deviation, and limits of agreement of respiration rate from acoustic monitoring versus capnography were 0.2 ± 2.3 breaths per min (bpm) and -4.2–4.7 bpm (Fig. 1), indicating good overall agreement in respiration rate estimations by the two methods. As compared with capnography, RRa had a mean difference of 2 bpm for data points \10 bpm (n = 22 188; P \ 0.0001,), no difference for data points between 10 and 20 bpm (n = 110 954), and a mean difference of -2 bpm for data points [20 bpm (n = 4 057; P \ 0.0001). Retrospective analysis validated 51 respiratory pauses from 13 patients (65 % of patients). These patients experienced between 1 and 11 respiratory pauses during their procedures. Of patients who experienced respiratory pauses, they tended to have either experienced one (n = 6) or more than 4 (n = 6) events. Two events (4 %) were detected by all three methods and 22 events (43 %) were detected by both acoustic monitoring and capnography. True positives, false positives, sensitivity and specificity detected by each method are shown in Table 2. The CI indicate that when compared with the other methods, the differences in sensitivity and specificity for any test method were significant. Clinician observation missed the most events, whereas capnography detected the most events but had many false positives. The LR? and LR- (upper and lower boundaries) for detection of respiratory pause was 3.5 (0.97–12.54) and 0.9 (0.8–1.0) for clinician observation, 1.0 (0.8–1.0) and 1.6 (0.5–4.9) for capnography, and -0.5 (0.4–0.7) and 4.1 (2.1–7.9) for acoustic monitoring. There was no relationship between respiration rate by acoustic monitoring and depth of sedation as indicated by PSI values (R2 = 0.0057). In patients who experienced respiratory pauses, there was no correlation between the depth of sedation as measured by the PSI and the incidence of respiratory pause. All episodes of respiratory pause occurred when the PSI was [40 (71.5 ± 15.4), except for one event in which the PSI was zero, which was considered
Table 1 Patient demographics
3 Results A total of 21 patients were initially enrolled in the study. One patient was excluded after consent because general anesthesia was required. Of the 20 included patients, there were 15 females and five males with a mean age of 61 years (range 32–87). Three patients required intubation during the case, so they contributed only a partial data set. The surgical procedures performed were total knee arthroplasties (18), tumor bone resection (1) and wrist reduction (1). Patient demographics are shown in Table 1.
N
20
Age (range), years
61.3 ± 14.5 (32–87)
Gender: female/male, n
15/5
Height, in.
64.6 ± 4.5
Weight, kg
85 ± 23
Procedures Knee replacement surgery
18
Tumor bone resection
1
Wrist open reduction
1
Values reported as number and mean ± standard deviation
123
J Clin Monit Comput Fig. 1 Bland Altman density plot of bias (red line) and limits of agreement (dashed line) comparing difference of respiration rate by acoustic monitoring (RRa) to the average respiration rate from RRa and capnography in breaths per minute (bpm)
Table 2 Sensitivity and specificity of three methods for detection of 51 respiratory pauses from 13 patients under sedation True positive, n Clinician observation
8
True negative, n
False positive, n
64
3
False negative, n
Sensitivity (CI), %
Specificity (CI), %
43
16 (7–28)
95 (87–99)
EtCO2
45
5
62
6
88 (76–95)
7 (2–16)
RRa
28
58
9
23
55 (40–68)
87 (76–93)
4 Discussion
Fig. 2 Distribution of patient state index (PSI) values from SEDLine brain function monitor during 50 validated respiratory pause events
an erroneous reading and was not included in this analysis (Fig. 2). Nine patients showed burst suppression on the SEDLine monitor, indicating periods of very deep sedation, but none of these periods was associated with the occurrence of respiratory pause (data not shown).
123
This study addresses the accuracy of detecting respiratory pauses by capnography and acoustic monitoring in patients undergoing regional anesthesia and MAC. Our series indicated that acoustic monitoring provided similar respiration rate values as capnography within 2 bpm in these patients, confirming what has been found in other studies that examined the accuracy of acoustic monitoring and capnography in post anesthesia care patients [6, 7] and those undergoing procedural sedation [8]. Ramsay et al. identified 21 respiratory pauses, defined as a cessation in breathing for C30 s, in 33 patients and found acoustic monitoring detected 81 % of events, whereas capnography detected only 62 % of these events, a difference that was significant. Our study, which defined a respiratory pause as cessation of breathing for C20 s, showed that capnography detected more events but had many more false alarms, a metric not examined by Ramsay et al. It is unclear what factors may have contributed to the relative differences in respiratory pause detection between the two studies, be it the different patient populations, the different definition
J Clin Monit Comput
used for respiratory pause, or some other factor. In another study, Goudra et al. [8] examined the accuracy of detection of respiratory pause, defined as a respiration rate of zero for C30 s, by acoustic monitoring and capnography during conscious sedation in 98 upper gastrointestinal endoscopy patients. Eleven true respiratory pause events were observed. Sensitivity and specificity for detection of respiratory pause were 73 and 93 % for acoustic monitoring and 73 and 12 % for capnography. Therefore, in this study acoustic monitoring and capnography showed a similar ability to detect respiratory pauses but, consistent with our study, capnography had far more false alarms (100 false positives as compared with eight for acoustic monitoring). It should be noted that Goudra et al. used clinical observation to verify respiratory pause events, which, in our study and others [3], was a poor means of detecting cessation of breathing. The merits and limitations of the aforementioned methods brings to the forefront the need for a single respiratory rate monitoring device that will capitalize on the strengths of existing non-invasive monitoring technologies while minimizing their weaknesses. One respiratory rate monitoring method not evaluated here is respiratory impedance pneumography. This technique is still limited by its susceptibility to motion and change in body position for instances. In the context of this study, combining the gas exchange information from capnography with the innovative acoustic monitoring could potentially improve events detection while reducing false alarms. While such a combination does not currently exist in a single machine, a recently released patient monitoring and connectivity platform called Root (Masimo, Irvine) may provide a useful alternative by integrating multiple streams of data (CO2, acoustic monitoring, and SedLine) in a single display for clinicians. Its effectiveness in clinical settings similar to that of this study has yet to be determined. Our study evaluated three methods for detecting respiratory pause during sedation, all of which have limitations. Clinician observation tends to miss many events, whereas capnography detects most events but tends to have many false alarms. Acoustic monitoring detected significantly more events than clinical observations but significantly fewer than capnography; however acoustic monitoring had significantly fewer false alarms as compared with capnography but more than clinician observation. Missed events endanger the patient because interventions are not triggered to rescue the patient. False alarms endanger the patient because clinicians may ignore the alarm if it is usually false, a phenomena referred to as alarm fatigue [10]. Because sensitivity and specificity are important in evaluating a device to detect a critical event, we used likelihood ratios, which are measures that combine both sensitivity and specificity into a single metric that indicates
the probability of a device detecting a true event. RRa had the highest likelihood ratio positive and the lowest likelihood ratio negative of the three test methods, indicating that this method has the combined highest probability of detecting true events with the lowest probability that an event will be missed. Therefore RRa may be the best method of those tested for monitoring ventilation during sedation. Patients develop apneic episodes when they are provided sedation or analgesic drugs. To avoid the morbidity and mortality of apnea, accurate continuous monitoring of respiratory rate seems appropriate. We hypothesized that the incidence of respiratory pause would increase as the level of sedation increased, because other studies have shown that depth of sedation is associated with the incidence of oxygen desaturations and airway obstructions in children [11–13]. However in our study we observed that cessation of breathing commonly occurred at moderate levels of sedation and that there were no episodes of respiratory pause at very deep levels of sedation (burst suppression on the observed EEG), but the number of burst suppression events we observed was small. We found that both apnea and excessive sedation occur in patients undergoing monitored anesthetic care (MAC) but do not occur at the same time. This suggests that both respiration rate monitoring for the detection of hypoventilation and brain function monitoring for depth of sedation are indicated for patients being provided sedation, even though respiratory compromise and excessive sedation may not be linked. There are however higher rates of complications with higher levels of sedation [14, 15]. For patients under MAC, precise monitoring of the depth of sedation facilitates the optimization of anesthetic dose and may prevent adverse complications such as desaturation, bradycardia, tachycardia, and hypotension during the procedure [16] and cognitive impairment after the procedure [17]. In our study, sedation was provided as clinically indicated, and was not guided by the brain function monitor. It would be a reasonable area of future study to determine whether less sedation would be given and possibly fewer episodes of respiratory pause observed (or other complications) if the brain function monitor and acoustic monitor were visible to the clinician. Our study has some limitations. Most of the patients (75 %) were female and underwent one type of procedure. Also, because this was a pilot study, only a small number of patients was evaluated. A larger study with more complex or longer procedures is needed to verify our findings. Second, we used a threshold of 20 s of cessation of breathing as a definition for respiratory pause. An episode of apnea, on the other hand, is defined by The American Academy of Sleep Medicine as an event that lasts for C10 s and is characterized by an absence of breathing with
123
J Clin Monit Comput
an oxygen desaturation of 3 % [18]. Therefore, our study did not formally measure the ability of these devices to detect apnea. Events of cessation of breathing that were less than 20 s were not counted, so any relative difference in the ability of one method to detect shorter events was not evaluated. We believe the 20-s threshold is appropriate for this application, however, because, unlike in sleep studies, the shorter episodes of cessation of breathing are probably not clinically relevant during sedation. Lastly, because we did not retrospectively review the data file for all the monitored time, we were not able to calculate a true sensitivity and specificity for detection of respiratory pause for each method. True positives were defined as those events that were determined to be true respiratory pauses of the events detected by any method, rather than the occurrence of any true event detected by the reference method (retrospective analysis). In conclusion, acoustic monitoring provided the highest probability of detecting respiratory pause and the lowest probability of missing a respiratory pause. When compared with clinical observation or capnography, acoustic monitoring of respiration rate may provide the best single method for monitoring ventilation during procedures requiring sedation because it has acceptable accuracy for detecting respiratory pauses, with a low rate of false alarms. Acknowledgments This study was supported by a Grant from Masimo Corporation, Irvine CA, USA. Conflict of interest of interest.
The authors declare that they have no conflict
Ethical standards All experiments conducted for this study comply with the current laws of the country in which they were performed (United States).
References 1. Krauss BGS. Procedural sedation and analgesia in children. Lancet. 2006;367(9512):766–80. 2. Weaver J. The latest ASA mandate CO2 monitoring for moderate and deep sedation. Anesth Prog. 2011;58(3):111–2. 3. Vargo JJ, Zuccaro G Jr, Dumot JA, Conwell DL, Morrow JB, Shay SS. Automated graphic assessment of respiratory activity is superior to pulse oximetry and visual assessment for the detection of early respiratory depression during therapeutic upper endoscopy. Gastrointest Endosc. 2002;55(7):826–31.
123
4. Fu ES, Downs JB, Schweiger JW, Miguel RV, Smith RA. Supplemental oxygen impairs detection of hypoventilation by pulse oximetry. Chest. 2004;126(5):1552–8. doi:10.1378/chest.126.5. 1552. 5. Maddox RR, Williams CK. Clinical experience with Capnography monitoring for PCA patients. APSF Newsletter. 2012;26:3. 6. Ramsay M.A.E. UM, Lagow E., Mendoza M., Untalan E., De Vol, E. The accuracy, precision and reliability of measuring ventilatory rate and detecting ventilatory pause by rainbow acoustic monitoring and capnometry. Anesth Analg. 2013;117(1): 69–75. 7. Mimoz O, Benard T, Gaucher A, Frasca D, Debaene B. Accuracy of respiratory rate monitoring using a non-invasive acoustic method after general anaesthesia. Br J Anaesth. 2012;108(5): 872–5. doi:10.1093/bja/aer510. 8. Goudra BG PL, Speck RM, Sinha AC. Comparison of acoustic respiration rate, impedance pneumography and capnometry monitoris for respiration rate accuracy and apnea detection during GI endoscopy. Open J Anesth. 2013;3:74–9. 9. Bland JM, Altman DG. Agreement between methods of measurement with multiple observations per individual. J Biopharm Stat. 2007;17(4):571–82. 10. Mitka M Joint commission warns of alarm fatigue: multitude of alarms from monitoring devices problematic. JAMA. 2013;309(22): 2315–16. doi:10.1001/jama.2013.6032. 11. Motas D, McDermott NB, VanSickle T, Friesen RH. Depth of consciousness and deep sedation attained in children as administered by nonanaesthesiologists in a children’s hospital. Paediatr Anaesth. 2004;14(3):256–60. 12. Malviya S, Voepel-Lewis T, Eldevik OP, Rockwell DT, Wong JH, Tait AR. Sedation and general anaesthesia in children undergoing MRI and CT: adverse events and outcomes. Br J Anaesth. 2000;84(6):743–8. 13. Malviya S, Voepel-Lewis T, Tait AR. Adverse events and risk factors associated with the sedation of children by nonanesthesiologists. Anesth Analg. 1997;85(6):1207–13. 14. Dawson N, Dewar A, Gray A, Leal A Association between ASA grade and complication rate in patients receiving procedural sedation for relocation of dislocated hip prostheses in a UK emergency department. Emerg Med J. 2013;1–3. doi: 10.1136/ emermed-2012-202147. 15. Jacques KG, Dewar A, Gray A, Kerslake D, Leal A, Lees F Procedural sedation and analgesia in a large UK Emergency Department: factors associated with complications. Emerg Med J. 2011;28:1036–40. doi:10.1136/emj.2010.102475. 16. Bateman ST, Lacroix J, Boven K, Forbes P, Barton R, Thomas NJ, Jacobs B, Markovitz B, Goldstein B, Hanson JH, Li HA, Randolph AG. Anemia, blood loss, and blood transfusions in North American children in the intensive care unit. Am J Respir Crit Care Med. 2008;178(1):26–33. doi:10.1164/rccm.2007111637OC. 17. Hughes CG, Pandharipande PP Review articles: the effects of perioperative and intensive care unit sedation on brain organ dysfunction. Anesth Analg. 2011;112(5):1212–7. doi:10.1213/ ANE.0b013e318215366d. 18. De Backer W. Obstructive sleep apnea/hypopnea syndrome. Panminerva Med. 2013;55(2):191–5.
ORIGINAL RESEARCH
Detection of respiratory compromise by acoustic monitoring, capnography, and brain function monitoring during monitored anesthesia care Pedro P. Tanaka • Maria Tanaka • David R. Drover
Received: 30 June 2013 / Accepted: 7 January 2014 Ó Springer Science+Business Media New York 2014
Abstract Episodes of apnea in sedated patients represent a risk of respiratory compromise. We hypothesized that acoustic monitoring would be equivalent to capnography for detection of respiratory pauses, with fewer false alarms. In addition, we hypothesized that the patient state index (PSI) would be correlated with the frequency of respiratory pauses and therefore could provide information about the risk of apnea during sedation. Patients undergoing sedation for surgical procedures were monitored for respiration rate using acoustic monitoring and capnography and for depth of sedation using the PSI. A clinician blinded to the acoustic and sedation monitor observed the capnograph and patient to assess sedation and episodes of apnea. Another clinician retrospectively reviewed the capnography and acoustic waveform and sound files to identify true positive and false positive respiratory pauses by each method (reference method). Sensitivity, specificity, and likelihood ratio for detection of respiratory pause was calculated for acoustic monitoring and capnography. The correlation of PSI with respiratory pause events was determined. For the 51 respiratory pauses validated by retrospective analysis, the sensitivity, specificity, and likelihood ratio positive for detection were 16, 96 %, and 3.5 for clinician observation; 88, 7 %, and 1.0 for capnography; and 55, 87 %, and 4.1 for acoustic monitoring. There was no correlation between PSI and respiratory pause events. Acoustic monitoring had the highest likelihood ratio positive for detection of respiratory pause events compared with capnography and clinician observation and,
P. P. Tanaka (&) M. Tanaka D. R. Drover Department of Anesthesiology, Perioperative and Pain Medicine, Stanford University School of Medicine, 300 Pasteur Dr. H3577, Stanford, CA 94305-5640, USA e-mail: [email protected]
therefore, may provide the best method for respiration rate monitoring during these procedures. Keywords Apnea Capnography Monitoring: intraoperative Sedation
1 Introduction The number of procedures being performed under sedation, and commonly under very deep sedation, is increasing [1]. In many situations, administering sedative drugs and monitoring the patient is not done by an anesthesiologist, but by less trained clinical personnel. Even when an anesthesiologist performs deep sedation, there are a variety of reasons why he or she may be remote from the airway, making it difficult to assess the level of sedation and respiratory effort. Patient monitoring that allows for the timely detection of hypoventilation is necessary during these procedures. The American Society of Anesthesiologists has amended its Standards for Basic Anesthetic Monitoring to include mandatory exhaled end-tidal carbon dioxide (ETCO2) monitoring during both moderate and deep sedation [2]. Clinician observation, pulse oximetry, and capnography are used individually or in combination to monitor ventilation during sedation, but each method has known limitations. Clinician observation is unreliable owing to competing priorities during procedures [3], pulse oximetry may be a late indicator of hypoventilation especially when supplemental oxygen is administered [4], and capnography can have a high rate of false alarms [5]. Rainbow Acoustic MonitoringTM by Pulse CO-Oximetry is a relatively new method of monitoring ventilation, and has been shown to provide accurate estimations of respiration rate in the post
123
J Clin Monit Comput
anesthesia care unit [6, 7] and during procedural sedation for endoscopy [8]. The purpose of this study was to compare the accuracy of acoustic monitoring and capnography for detection of respiratory pauses during sedation as compared with a retrospective analysis of waveform and sound files analyzed by a clinician. As a secondary endpoint we examined whether apnea is associated with the level of sedation as determined by patient state index (PSI) during monitored anesthesia care (MAC). We hypothesized that during deep sedation it would be more likely for the airway to become obstructed and so tested whether the depth of sedation was associated with respiratory pause events.
2 Methods After receiving approval by the Stanford University Institutional Review board and written informed consent of patients, we conducted a prospective, single-center study of patients ASA I to III, 18 years of age or older, who were scheduled to undergo sedation with local or regional anesthesia. To facilitate insertion of the local or regional block, all patients received premedication consisting of midazolam and fentanyl administered intravenously (IV) at the discretion of the anesthesiologist or regional anesthesia team in the preoperative area. All patients received standard monitoring consisting of oxygen saturation by pulse oximetry (SpO2), noninvasive blood pressure measurement, end-tidal carbon dioxide monitoring (capnography, ETCO2) using tubing with a standard nasal cannula, skin temperature, and electroencephalogram recordings. An additional capnography monitor (CapnostreamÒ 20, Oridion Capnography, Inc., Bedford, MA, USA) was connected to the same nasal cannula to collect study data. Oxygen was administered continuously through a facial mask at a rate of 4–10 L/min. During the procedure, repositioning of the patient’s head, insertion of a nasal or oral airway and decreasing or increasing sedation was performed at the discretion of the anesthesiologist and based on the standard of care protocol. An adhesive acoustic respiration sensor (RAS-125TM revision C) was placed on the patient’s necks and connected to a Pulse COOximeter with Rainbow Acoustic Monitoring (Rad-87, software v. 7805, Masimo Corp., Irvine CA, USA) to provide acoustic respiration rate (RRa), SpO2, pulse rate (PR), and perfusion index (PI). A stethoscope was attached to the patient’s neck on the side contralateral to the adhesive acoustic sensor. A sensor array connected to a brain function monitor (SEDLine, Masimo Corp, Irvine CA, USA) was applied to the forehead of all patients to monitor the level of sedation, as indicated by the PSI. The capnograph, brain function monitor, and acoustic monitor data
123
were all electronically synchronized and connected to a laptop computer running data collection software (automatic data collection, Masimo Corporation) for continuous data recording. In patients who presented with a flat carbon dioxide line on the standard of care capnograph for 10 s, absence of airflow was confirmed by listening to the neck with the stethoscope by a clinician blinded to the acoustic monitor, study capnograph and SEDLine monitor. These events were recorded as respiratory pauses. A second clinician not blinded to the study monitors marked in the electronic record any respiratory pause event detected by each method. For the purposes of the study, a respiratory pause was defined as a cessation of breathing for C20 s. After the procedure, a clinician replayed the waveform and sound files recorded from the acoustic monitor and capnograph using a specialized LabView-based software program (Tag Editor, Masimo) to determine the validity of the respiratory pauses detected by each method. Tag Editor allows for the simultaneous viewing of both the acoustic and capnography waveforms while listening to the breathing sounds from the acoustic signal, to determine inspiration and expiration reference markers within the respiration cycle independently from the acoustic monitor or capnometercalculated respiration rates. This method has been used in another study investigating the accuracy of acoustic monitoring for detection of respiratory pauses [6]. Periods of approximately 1 min before to 1 min after each presumed respiratory pause were reviewed. Therefore, any respiratory pause event missed by all methods was not included in the analysis. Periods of respiratory pause were determined by (1) clinical judgment, (2) respiratory rate as determined by EtCO22 (RR ETCO2), and (3) RRa. If a patient required intubation based on standard of care, data collection was stopped for that patient but the data collected up to that point contributed to the data set. 2.1 Data analysis A power analysis was not feasible because there were no preliminary data on which to base the analysis. Thus, we deemed this to be a pilot study. To evaluate the agreement between capnography and acoustic monitoring for estimation of respiration rate, bias (mean of differences between methods), standard deviation of the bias, and limit of agreement were calculated and graphed using the Bland–Altman method [9]. To determine the accuracy of detection of respiratory pauses, we calculated the number of respiratory pauses detected that were validated (true positives), the number of respiratory pauses detected and not validated (false positives), and the number of true positive detected by each test method divided by the total number of true events.
J Clin Monit Comput
To determine the sensitivity and specificity for each method, we used the following definitions: A true positive was a presumed respiratory pause detected by a test method that was verified by retrospective analysis. A true negative was no respiratory pause detected by the test method or the reference method at a time when one or more other test methods detected a respiratory pause. A false positive was a presumed respiratory pause detected by a test method that was not validated by retrospective analysis. A false negative was a respiratory pause verified by retrospective analysis that was not detected by the test method. Sensitivity, therefore, was equal to the number of true positives divided by the number of true positives plus the number of false negatives. Specificity was equal to the number of true negatives divided by the number of true negatives plus the number of false positives. Confidence intervals (CI) for sensitivity and specificity for each test method were calculated by a bootstrapping technique, where a resampling with replacement was used with 1,000 samples from 118 available data points to provide an empirically derived distribution to estimate the upper and lower boundaries. Likelihood ratios were used to provide a single metric that combines sensitivity and specificity to compare the test methods. For the purposes of this study, a likelihood ratio positive (LR?) indicates the probability that the detected respiratory pause was a true event, with a larger number indicating a higher probability that a respiratory pause would be detected by the method. In contrast, a likelihood ratio negative (LR-) is the probability of a true respiratory pause not being detected by the test method, with the smaller number indicating a lower probability of a missed event. Likelihood ratios with upper and lower boundaries were calculated for each test method. To determine if there was a relationship between respiration rate and depth of sedation, a regression plot of RRa values to PSI values was plotted and the correlation of determination (R2) was calculated. Additionally, the distribution of PSI values verses the frequency of respiratory pauses was plotted.
A total of 137,199 paired ETCO2 and RRa respiration rate values were collected. The bias, standard deviation, and limits of agreement of respiration rate from acoustic monitoring versus capnography were 0.2 ± 2.3 breaths per min (bpm) and -4.2–4.7 bpm (Fig. 1), indicating good overall agreement in respiration rate estimations by the two methods. As compared with capnography, RRa had a mean difference of 2 bpm for data points \10 bpm (n = 22 188; P \ 0.0001,), no difference for data points between 10 and 20 bpm (n = 110 954), and a mean difference of -2 bpm for data points [20 bpm (n = 4 057; P \ 0.0001). Retrospective analysis validated 51 respiratory pauses from 13 patients (65 % of patients). These patients experienced between 1 and 11 respiratory pauses during their procedures. Of patients who experienced respiratory pauses, they tended to have either experienced one (n = 6) or more than 4 (n = 6) events. Two events (4 %) were detected by all three methods and 22 events (43 %) were detected by both acoustic monitoring and capnography. True positives, false positives, sensitivity and specificity detected by each method are shown in Table 2. The CI indicate that when compared with the other methods, the differences in sensitivity and specificity for any test method were significant. Clinician observation missed the most events, whereas capnography detected the most events but had many false positives. The LR? and LR- (upper and lower boundaries) for detection of respiratory pause was 3.5 (0.97–12.54) and 0.9 (0.8–1.0) for clinician observation, 1.0 (0.8–1.0) and 1.6 (0.5–4.9) for capnography, and -0.5 (0.4–0.7) and 4.1 (2.1–7.9) for acoustic monitoring. There was no relationship between respiration rate by acoustic monitoring and depth of sedation as indicated by PSI values (R2 = 0.0057). In patients who experienced respiratory pauses, there was no correlation between the depth of sedation as measured by the PSI and the incidence of respiratory pause. All episodes of respiratory pause occurred when the PSI was [40 (71.5 ± 15.4), except for one event in which the PSI was zero, which was considered
Table 1 Patient demographics
3 Results A total of 21 patients were initially enrolled in the study. One patient was excluded after consent because general anesthesia was required. Of the 20 included patients, there were 15 females and five males with a mean age of 61 years (range 32–87). Three patients required intubation during the case, so they contributed only a partial data set. The surgical procedures performed were total knee arthroplasties (18), tumor bone resection (1) and wrist reduction (1). Patient demographics are shown in Table 1.
N
20
Age (range), years
61.3 ± 14.5 (32–87)
Gender: female/male, n
15/5
Height, in.
64.6 ± 4.5
Weight, kg
85 ± 23
Procedures Knee replacement surgery
18
Tumor bone resection
1
Wrist open reduction
1
Values reported as number and mean ± standard deviation
123
J Clin Monit Comput Fig. 1 Bland Altman density plot of bias (red line) and limits of agreement (dashed line) comparing difference of respiration rate by acoustic monitoring (RRa) to the average respiration rate from RRa and capnography in breaths per minute (bpm)
Table 2 Sensitivity and specificity of three methods for detection of 51 respiratory pauses from 13 patients under sedation True positive, n Clinician observation
8
True negative, n
False positive, n
64
3
False negative, n
Sensitivity (CI), %
Specificity (CI), %
43
16 (7–28)
95 (87–99)
EtCO2
45
5
62
6
88 (76–95)
7 (2–16)
RRa
28
58
9
23
55 (40–68)
87 (76–93)
4 Discussion
Fig. 2 Distribution of patient state index (PSI) values from SEDLine brain function monitor during 50 validated respiratory pause events
an erroneous reading and was not included in this analysis (Fig. 2). Nine patients showed burst suppression on the SEDLine monitor, indicating periods of very deep sedation, but none of these periods was associated with the occurrence of respiratory pause (data not shown).
123
This study addresses the accuracy of detecting respiratory pauses by capnography and acoustic monitoring in patients undergoing regional anesthesia and MAC. Our series indicated that acoustic monitoring provided similar respiration rate values as capnography within 2 bpm in these patients, confirming what has been found in other studies that examined the accuracy of acoustic monitoring and capnography in post anesthesia care patients [6, 7] and those undergoing procedural sedation [8]. Ramsay et al. identified 21 respiratory pauses, defined as a cessation in breathing for C30 s, in 33 patients and found acoustic monitoring detected 81 % of events, whereas capnography detected only 62 % of these events, a difference that was significant. Our study, which defined a respiratory pause as cessation of breathing for C20 s, showed that capnography detected more events but had many more false alarms, a metric not examined by Ramsay et al. It is unclear what factors may have contributed to the relative differences in respiratory pause detection between the two studies, be it the different patient populations, the different definition
J Clin Monit Comput
used for respiratory pause, or some other factor. In another study, Goudra et al. [8] examined the accuracy of detection of respiratory pause, defined as a respiration rate of zero for C30 s, by acoustic monitoring and capnography during conscious sedation in 98 upper gastrointestinal endoscopy patients. Eleven true respiratory pause events were observed. Sensitivity and specificity for detection of respiratory pause were 73 and 93 % for acoustic monitoring and 73 and 12 % for capnography. Therefore, in this study acoustic monitoring and capnography showed a similar ability to detect respiratory pauses but, consistent with our study, capnography had far more false alarms (100 false positives as compared with eight for acoustic monitoring). It should be noted that Goudra et al. used clinical observation to verify respiratory pause events, which, in our study and others [3], was a poor means of detecting cessation of breathing. The merits and limitations of the aforementioned methods brings to the forefront the need for a single respiratory rate monitoring device that will capitalize on the strengths of existing non-invasive monitoring technologies while minimizing their weaknesses. One respiratory rate monitoring method not evaluated here is respiratory impedance pneumography. This technique is still limited by its susceptibility to motion and change in body position for instances. In the context of this study, combining the gas exchange information from capnography with the innovative acoustic monitoring could potentially improve events detection while reducing false alarms. While such a combination does not currently exist in a single machine, a recently released patient monitoring and connectivity platform called Root (Masimo, Irvine) may provide a useful alternative by integrating multiple streams of data (CO2, acoustic monitoring, and SedLine) in a single display for clinicians. Its effectiveness in clinical settings similar to that of this study has yet to be determined. Our study evaluated three methods for detecting respiratory pause during sedation, all of which have limitations. Clinician observation tends to miss many events, whereas capnography detects most events but tends to have many false alarms. Acoustic monitoring detected significantly more events than clinical observations but significantly fewer than capnography; however acoustic monitoring had significantly fewer false alarms as compared with capnography but more than clinician observation. Missed events endanger the patient because interventions are not triggered to rescue the patient. False alarms endanger the patient because clinicians may ignore the alarm if it is usually false, a phenomena referred to as alarm fatigue [10]. Because sensitivity and specificity are important in evaluating a device to detect a critical event, we used likelihood ratios, which are measures that combine both sensitivity and specificity into a single metric that indicates
the probability of a device detecting a true event. RRa had the highest likelihood ratio positive and the lowest likelihood ratio negative of the three test methods, indicating that this method has the combined highest probability of detecting true events with the lowest probability that an event will be missed. Therefore RRa may be the best method of those tested for monitoring ventilation during sedation. Patients develop apneic episodes when they are provided sedation or analgesic drugs. To avoid the morbidity and mortality of apnea, accurate continuous monitoring of respiratory rate seems appropriate. We hypothesized that the incidence of respiratory pause would increase as the level of sedation increased, because other studies have shown that depth of sedation is associated with the incidence of oxygen desaturations and airway obstructions in children [11–13]. However in our study we observed that cessation of breathing commonly occurred at moderate levels of sedation and that there were no episodes of respiratory pause at very deep levels of sedation (burst suppression on the observed EEG), but the number of burst suppression events we observed was small. We found that both apnea and excessive sedation occur in patients undergoing monitored anesthetic care (MAC) but do not occur at the same time. This suggests that both respiration rate monitoring for the detection of hypoventilation and brain function monitoring for depth of sedation are indicated for patients being provided sedation, even though respiratory compromise and excessive sedation may not be linked. There are however higher rates of complications with higher levels of sedation [14, 15]. For patients under MAC, precise monitoring of the depth of sedation facilitates the optimization of anesthetic dose and may prevent adverse complications such as desaturation, bradycardia, tachycardia, and hypotension during the procedure [16] and cognitive impairment after the procedure [17]. In our study, sedation was provided as clinically indicated, and was not guided by the brain function monitor. It would be a reasonable area of future study to determine whether less sedation would be given and possibly fewer episodes of respiratory pause observed (or other complications) if the brain function monitor and acoustic monitor were visible to the clinician. Our study has some limitations. Most of the patients (75 %) were female and underwent one type of procedure. Also, because this was a pilot study, only a small number of patients was evaluated. A larger study with more complex or longer procedures is needed to verify our findings. Second, we used a threshold of 20 s of cessation of breathing as a definition for respiratory pause. An episode of apnea, on the other hand, is defined by The American Academy of Sleep Medicine as an event that lasts for C10 s and is characterized by an absence of breathing with
123
J Clin Monit Comput
an oxygen desaturation of 3 % [18]. Therefore, our study did not formally measure the ability of these devices to detect apnea. Events of cessation of breathing that were less than 20 s were not counted, so any relative difference in the ability of one method to detect shorter events was not evaluated. We believe the 20-s threshold is appropriate for this application, however, because, unlike in sleep studies, the shorter episodes of cessation of breathing are probably not clinically relevant during sedation. Lastly, because we did not retrospectively review the data file for all the monitored time, we were not able to calculate a true sensitivity and specificity for detection of respiratory pause for each method. True positives were defined as those events that were determined to be true respiratory pauses of the events detected by any method, rather than the occurrence of any true event detected by the reference method (retrospective analysis). In conclusion, acoustic monitoring provided the highest probability of detecting respiratory pause and the lowest probability of missing a respiratory pause. When compared with clinical observation or capnography, acoustic monitoring of respiration rate may provide the best single method for monitoring ventilation during procedures requiring sedation because it has acceptable accuracy for detecting respiratory pauses, with a low rate of false alarms. Acknowledgments This study was supported by a Grant from Masimo Corporation, Irvine CA, USA. Conflict of interest of interest.
The authors declare that they have no conflict
Ethical standards All experiments conducted for this study comply with the current laws of the country in which they were performed (United States).
References 1. Krauss BGS. Procedural sedation and analgesia in children. Lancet. 2006;367(9512):766–80. 2. Weaver J. The latest ASA mandate CO2 monitoring for moderate and deep sedation. Anesth Prog. 2011;58(3):111–2. 3. Vargo JJ, Zuccaro G Jr, Dumot JA, Conwell DL, Morrow JB, Shay SS. Automated graphic assessment of respiratory activity is superior to pulse oximetry and visual assessment for the detection of early respiratory depression during therapeutic upper endoscopy. Gastrointest Endosc. 2002;55(7):826–31.
123
4. Fu ES, Downs JB, Schweiger JW, Miguel RV, Smith RA. Supplemental oxygen impairs detection of hypoventilation by pulse oximetry. Chest. 2004;126(5):1552–8. doi:10.1378/chest.126.5. 1552. 5. Maddox RR, Williams CK. Clinical experience with Capnography monitoring for PCA patients. APSF Newsletter. 2012;26:3. 6. Ramsay M.A.E. UM, Lagow E., Mendoza M., Untalan E., De Vol, E. The accuracy, precision and reliability of measuring ventilatory rate and detecting ventilatory pause by rainbow acoustic monitoring and capnometry. Anesth Analg. 2013;117(1): 69–75. 7. Mimoz O, Benard T, Gaucher A, Frasca D, Debaene B. Accuracy of respiratory rate monitoring using a non-invasive acoustic method after general anaesthesia. Br J Anaesth. 2012;108(5): 872–5. doi:10.1093/bja/aer510. 8. Goudra BG PL, Speck RM, Sinha AC. Comparison of acoustic respiration rate, impedance pneumography and capnometry monitoris for respiration rate accuracy and apnea detection during GI endoscopy. Open J Anesth. 2013;3:74–9. 9. Bland JM, Altman DG. Agreement between methods of measurement with multiple observations per individual. J Biopharm Stat. 2007;17(4):571–82. 10. Mitka M Joint commission warns of alarm fatigue: multitude of alarms from monitoring devices problematic. JAMA. 2013;309(22): 2315–16. doi:10.1001/jama.2013.6032. 11. Motas D, McDermott NB, VanSickle T, Friesen RH. Depth of consciousness and deep sedation attained in children as administered by nonanaesthesiologists in a children’s hospital. Paediatr Anaesth. 2004;14(3):256–60. 12. Malviya S, Voepel-Lewis T, Eldevik OP, Rockwell DT, Wong JH, Tait AR. Sedation and general anaesthesia in children undergoing MRI and CT: adverse events and outcomes. Br J Anaesth. 2000;84(6):743–8. 13. Malviya S, Voepel-Lewis T, Tait AR. Adverse events and risk factors associated with the sedation of children by nonanesthesiologists. Anesth Analg. 1997;85(6):1207–13. 14. Dawson N, Dewar A, Gray A, Leal A Association between ASA grade and complication rate in patients receiving procedural sedation for relocation of dislocated hip prostheses in a UK emergency department. Emerg Med J. 2013;1–3. doi: 10.1136/ emermed-2012-202147. 15. Jacques KG, Dewar A, Gray A, Kerslake D, Leal A, Lees F Procedural sedation and analgesia in a large UK Emergency Department: factors associated with complications. Emerg Med J. 2011;28:1036–40. doi:10.1136/emj.2010.102475. 16. Bateman ST, Lacroix J, Boven K, Forbes P, Barton R, Thomas NJ, Jacobs B, Markovitz B, Goldstein B, Hanson JH, Li HA, Randolph AG. Anemia, blood loss, and blood transfusions in North American children in the intensive care unit. Am J Respir Crit Care Med. 2008;178(1):26–33. doi:10.1164/rccm.2007111637OC. 17. Hughes CG, Pandharipande PP Review articles: the effects of perioperative and intensive care unit sedation on brain organ dysfunction. Anesth Analg. 2011;112(5):1212–7. doi:10.1213/ ANE.0b013e318215366d. 18. De Backer W. Obstructive sleep apnea/hypopnea syndrome. Panminerva Med. 2013;55(2):191–5.
