Frequency and Temporal Distribution of Postoperative Respiratory Depressive Events : Anesthesia & Analgesia

Secondary Logo

Journal Logo

Featured Articles: Original Clinical Research Report

Frequency and Temporal Distribution of Postoperative Respiratory Depressive Events

Driver, C. Noelle BS*; Laporta, Mariana L. MD*; Bergese, Sergio D. MD; Urman, Richard D. MD; Di Piazza, Fabio MS§; Overdyk, Frank J. MD; Sprung, Juraj MD, PhD*; Weingarten, Toby N. MD*

Author Information
doi: 10.1213/ANE.0000000000005478



  • Question: What is the temporal distribution of postoperative respiratory depressive episodes as determined by continuous bedside capnography and pulse oximetry monitoring?
  • Findings: Postoperative respiratory depressive episodes are common, frequently multiple, with initial episodes typically occurring in the afternoon to early evening and rate of all episodes peaking in the early morning.
  • Meaning: Postoperative monitoring strategies should consider both the high prevalence and temporal distribution of respiratory depressive episodes.

Postoperative respiratory depression (RD) can lead to respiratory failure resulting in severe morbidity and mortality.1 Whereas previous reports suggest postoperative RD episodes were uncommon,2 recent investigations suggest the incidence may be greater than 40% and often go unrecognized.3–5 Preoperative recognition of patients at increased risk for postoperative RD can allow for implantation of mitigating perioperative measures such as more advanced postoperative monitoring. Recently, The PRediction of Opioid-induced respiratory Depression In patients monitored by capnoGraphY (PRODIGY) trial was conducted to create a multivariable risk prediction tool to classify patients as low, intermediate, or high risk for RD.3 In the PRODIGY trial, continuous capnography and pulse oximetry collected on general care wards were used to identify RD episodes.

Despite advances in assessing risk for postoperative RD, evidence as to when RD events occur is incomplete and conflicting. Cases of patients with sleep-disordered breathing found “dead in bed” from unrecognized nocturnal respiratory arrests understandably garner attention.6 Further polysomnography on surgical patients shows sleep-disordered breathing worsens postoperatively,7 which could lead to nocturnal fatal events. However, retrospective studies of postoperative naloxone administration to treat severe opioid-related RD find a substantial proportion of events occur in the first few hours following postanesthesia care unit (PACU) discharge.8–10 Determining when postoperative RD events occur has practical implications in devising monitoring strategies for high-risk patients.

Because the aim of the PRODIGY trial was to devise a risk stratification tool for which patients develop RD, data for each patient were analyzed only until the initial RD episode was identified. Thus, a large amount of data from the PRODIGY trial remains unanalyzed. In this study, data from 2 participating US centers were analyzed to identify subsequent RD episodes and the timing of RD events vis-à-vis end of surgery and time of day. The objective of this study was to describe the temporal pattern of these RD episodes (both initial and subsequent RD episodes) to determine if they predominately occur in the first hours following PACU discharge or nocturnally. Because during the initial PRODIGY analysis it was noted, informally, that many patients had multiple RD episodes, secondary study objectives were to quantify how many RD episodes occur in patients and if associations exist between the PRODIGY risk category and the frequency of postoperative RD episodes.


PRODIGY (prospective trial conducted between April 2017 and May 2018 at 16 sites in 7 countries) was approved by the appropriate institutional review board (IRB) and written informed consent was obtained from all subjects (subjects included adults [≥18, 20, and 21 years in the United States/Europe, Japan, and Singapore, respectively]) before enrollment.3 The trial was registered before patient enrollment at (NCT02811302, principal investigator: Frank J. Overdyk, date of registration: June 2016).

This study was a post hoc analysis of the PRODIGY trial dataset from patients enrolled at 2 US trial sites, approved by the IRB (Brigham and Women’s Hospital, Boston, MA [IRB #00010756, #00010757, #00010758, #00010759, #00010760, #00010761, #00010762] and the Ohio State University Wexner Medical Center, Columbus, OH [IRB #00000533]). The decision to limit this study to 2 sites was made for practical considerations, primarily the time commitment for confirming each RD episode from raw data. These 2 sites were selected because of the large number of enrolled patients and consistently high-quality data.

This article adheres to the applicable STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) guidelines.


PRODIGY3 was an observational trial in which surgical and medical patients receiving parenteral opioids on the general care floor were continuously monitored by pulse oximetry and capnography using the Capnostream 20p or 35 (Medtronic, Boulder, CO) for up to 48 hours.3,11 In that study, the Capnostream monitor was applied to patients in the general care ward and used to continuously collect expired carbon dioxide (Eco2), heart rate, and oxyhemoglobin saturation (Spo2) readings. Output from the monitor was blinded from the health care team and patients underwent standard postoperative monitoring, which at these institutions was via intermittent manual vital sign assessments. Eco2 values were used to calculate respiratory rate. These readings were subsequently interrogated by a computer algorithm for potential episodes of RD, which were defined as respiratory rate ≤5 breaths per minute (bpm) for ≥3 minutes; Spo2 ≤85% for ≥3 minutes; Eco2 ≤15 or ≥60 mm Hg for ≥3 minutes; apnea episode >30 seconds.3,11 These potential RD episodes were then manually reviewed and adjudicated as actual RD episode or artifact (Figure 1 is an example of collected capnography and oximetry data from a study subject which was deemed to be RD events). The primary trial objective was to derive a RD risk prediction tool from patient physiological and RD episode data (ie, PRODIGY risk prediction tool).3 The PRODIGY risk prediction tool uses using 5 clinical factors (age >60 in decades, male sex, sleep-disordered breathing, opioid naivety, and chronic heart failure), assigning patients to low, intermediate, or high risk for RD categories based on the final score (<8, ≥8 to <15, and ≥15, respectively).3

Figure 1.:
Repetitive postoperative apneic events as recorded by bedside capnography and pulse oximetry using the Capnostream monitor. The top panel displays exhaled carbon dioxide (blue) over time as measured by bedside capnography. The dashed blue lines indicate potential apneic respiratory depressive episodes. The bottom panel displays data from pulse oximetry oxyhemoglobin concentration (red) and heart rate (green). The repetitive decreases in oxyhemoglobin saturation and increases in heart rate are typically observed during this breathing pattern. Etco 2 indicates end-tidal carbon dioxide; PR, pulse rate; RR, respiratory rate; Spo2, oxyhemoglobin saturation.

Both centers selected for this post hoc analysis are major academic institutions in the United States. At both institutions nursing to patient ratio varies from 1:2 to 1:4, and shifts are typically 12 hours in duration (shift change at 7:00 and 19:00).


The main objective of this post hoc analysis was to characterize the frequency and temporal distribution of RD episodes after surgery. As part of this analysis, the time difference between the end of surgery and the occurrence of the first RD episode per patient was determined, and the frequency of RD episodes was evaluated by time of day. A secondary objective was to retrospectively determine the association between the number of RD episodes and PRODIGY risk score3 for each patient.

Patient Cohort

Patients from the 2 sites who were hospitalized for a surgical procedure, received postoperative parenteral opioids, and underwent continuous monitoring were included in the analysis cohort. Patients who were admitted to the intensive care unit before the general care floor or medical patients who did not undergo a surgical procedure were excluded. Sample size was determined without a priori power calculations.

RD Episode Adjudication and Data Collection

For reasons stated previously, the clinical event committee for the PRODIGY trial reviewed each patient’s data stream only until the first potential RD episode was manually verified, resulting in 2 patient groups: those with ≥1 RD episode, and those with no RD episodes.3 A “potential” episode of RD that met the threshold deviation criteria for the monitor specified above only became confirmed if the monitor tracings were consistent with physiologic and clinical findings in the record, thus eliminating RD episodes that may have been artifact due to patient movement, dislodging of the transducer, or other common reasons for artifact. In this post hoc analysis, however, all “potential” RD episodes for the entirety of the data stream for each patient were reviewed and verified (by authors C.N.D., M.L.L., T.N.W.) using the same criteria as in the initial study.3,11 For each patient, the number of RD episodes and time of each RD episode were recorded. A PRODIGY risk score was calculated for each subject using data collected during the original PRODIGY trial.3

Statistical Analysis

Categorical and continuous variables are described using counts and percentages, means and standard deviation, and median and interquartile ranges (IQR), respectively. The difference between proportions of patients who had RD episodes was tested by means of χ2 test, while for continuous variables, such as the time between the end of the surgery and the start of monitoring, length of monitoring and number of RD episodes per patient were analyzed using the Wilcoxon/Mann-Whitney test or Kruskal-Wallis test, when 2 or more groups were tested, respectively. The number of “initial” (first confirmed RD event for a subject) and “all” (initial and all subsequent) RD episodes were placed in 2-hour increments and displayed as 24-hour radar plots. These plots underwent visual inspection to determine peak concentration of events. The visual inspection of all RD episodes temporal distribution identified 2 peak occurrences (see Results section). To further assess if risk for all RD episodes differs by time intervals, a post hoc analysis was conducted using generalized estimating equations to construct models with the number of RD episodes as the dependent variable and the timing of the RD as independent variables accounting for nonindependent observations. Estimates were determined using a log link function and Poisson distribution. An autoregressive-1 correlation structure and robust standard errors were used. All statistical tests used a 2-sided significance level of 0.05. Statistical analysis was performed using SAS v9.4 (SAS Institute Inc, Cary, NC).


Patient Characteristics

Table 1. - Baseline Characteristics
Characteristics All patients (n = 250)
Age, y 53.2 ± 14.2
 <60 64.8% (162)
 ≥60 to <70 21.6% (54)
 ≥70 to <80 12.0% (30)
 ≥80 1.6% (4)
Surgery type
 General 63.2% (158)
 Neurosurgery 24.0% (60)
 Urologic-gynecologic 7.6% (19)
 Miscellaneous 5.2% (13)
Sex (male) 32.0% (80)
Body mass index, kg/m2 35.8 ± 11.4
Opioid naive 86.0% (215)
Chronic heart failurea 1.2% (3)
Sleep disorders 18.4% (46)
ASA physical status ≥III 73.6 (184)
Length of surgery, h
 <2 39.2% (98)
 ≥2 to <4 42.0% (105)
 ≥4 18.8% (47)
PRODIGY risk scorea
 Low 40.2% (100)
 Intermediate 31.7% (79)
 High 28.1% (70)
Data summarized as mean ± standard deviation or percentage (number).
Abbreviations: ASA, American Society of Anesthesiologists; PRODIGY, Prediction of Opioid-induced Respiratory Depression in Patients Monitored by Capnography.
a1 patient with missing information for 1 dimension (history of chronic heart failure) of the PRODIGY score.

This analysis includes data collected in a subset of patients (n = 274) enrolled in the PRODIGY study and belonging to 2 sites participating in the PRODIGY trial. Excluded were 24 patients: 16 due to a major protocol deviation, 5 were medical patients, and 3 were admitted to the general care floor from intensive care unit more than 24 hours after surgery. Within this set of patients who underwent a surgery, 248 were transferred to the inpatient ward from the PACU and 2 were admitted to the ward directly from the operative room. Table 1 summarizes baseline characteristics of these patients. There were substantial differences between subjects from these 2 sites and other sites both from the United States as well as the entire PRODIGY trial (Supplemental Digital Content 1, Table,

RD Episodes

The median time between the end of the surgery and the start of monitoring was 4.3 [3.2–6.2] hours and the median length of monitoring was 22.3 [12.9–24.3] hours, neither of which differed between PRODIGY risk categories (P = .67 and P = .14, the start and duration of monitoring, respectively). Review of Capnostream data identified 2539 RD episodes in 155 (62.0%, 95% confidence interval [CI], 55.7-68.0) patients with a median number of episodes per patient of 2 [0–8] and a range of 0–545 episodes per patient (Figure 1). Of those patients who had an RD episode, 136 (87.7%) had more than 1 event.

Median time of occurrence of the first confirmed (initial) episode of RD after surgery was 8.8 [5.1–18.0] postoperative hours, while the median time for any RD episodes was 15.7 [10.0–22.6] hours (Figure 2). Two-hundred twelve (84.8%) patients started the monitoring within 8 hours of surgery completion, while for 38 (15.2%) patients, monitoring started after 8 postsurgical hours (from 9 to 25 hours). A sensitivity analysis was performed on 212 patients who started the monitoring within 8 hours after surgery. One hundred and thirty-two (62.3%, 95% CI, 55.7-68.9) patients had at least 1 confirmed RD episode with a total of 2388 episodes, with a median time to first confirmed RD episode of 7.3 [4.8–12.8] hours and the median time for all confirmed RD episodes was 15.0 [9.8–21.1] hours.

Figure 2.:
The number of respiratory depressive episodes per patient by different PRODIGY risk scores. Box plot representation of the distribution of the number of RD episodes by PRODIGY score. The boxes indicate the IQR, with the midline representing the median and the diamond representing the mean. Whiskers extend from each box to the farthest point within ±1.5 × the IQR. Observations outside this range are considered outliers and denoted by small circles. IQR indicates interquartile range; PRODIGY, The PRediction of Opioid-induced respiratory Depression In patients monitored by capnoGraphY; RD, respiratory depression.

The temporal distributions for initial RD episodes and for all RD episodes are displayed as 24-hour radar plots in Figure 3. Initial RD episodes occurred frequently between 14:00 and 20:00 on the day of surgery. Visual inspection of the radar plot of all RD episodes suggested there were 2 peaks in incidence, an initial peak occurring between 16:00 and 22:00 the day of surgery, followed by a larger peak between 00:00 and 08:00 the morning after surgery. To confirm the visual inspection, a subsequent post hoc analysis was performed to assess risk of RD per time interval (Supplemental Digital Content 2, Figure, and found the 16:00–22:00 peak to be nonsignificant but that there was increased risk at 02:00, 04:00, and 06:00 (1.7 [1.1–2.6], P = .029; 2.1 [1.3–3.4], P = .004; and 1.8 [1.2–2.8], P = .009, respectively). The temporal distribution pattern was not affected by the exclusion of the subset of patients for whom monitoring was initiated greater than 8 hours after surgery.

Figure 3.:
Radar plots depicting time of day on 24-h clock and the number of respiratory depressive episodes. A, The time of day of the end of surgery and the initial postoperative respiratory depressive episodes. B, The time of day for all postoperative respiratory depressive episodes. The spokes on the radar plot represent the time of day on a 24-h clock. The magnitude of each spoke is the total number of episodes between the previous spoke time and the current spoke time (eg, at 02:00 is shown the number of episodes occurred between 00:00 and 02:00). The scale of episodes is different between the 2 plots. RD indicates respiratory depression.
Table 2. - PRODIGY Risk Score Category and Number of Postoperative Respiratory Events per Patienta
PRODIGY score Patients with ≥1 RD episodes
n (%) [95% CI]
Number RD episodes per patient
Median [IQR]
Number RD episodes per patient per hour
Median [IQR]
Low (n = 100) 47 (47.0%) [37.2-56.8] 0.0 [0–4]
0.0 [0–0.2]
Intermediate (n = 79) 49 (62.0%) [51.3-72.7] 1.0 [0–10]
0.1 [0–0.4]
High (n = 70) 59 (84.3%) [73.6-91.9] 5 [1–16]
0.3 [0.1–0.9]
Data presented as number (percentage) [95% CI], or median [IQR] minimum–maximum range.
Abbreviations: CI, confidence interval; IQR, interquartile range; PRODIGY, Prediction of Opioid-induced Respiratory Depression in Patients Monitored by Capnography; RD, respiratory depression.
aOverall comparison between groups for both incidence [χ2] and number of episodes [Kruskal-Wallis test]. All comparisons were statistically significant, P < .001.

Figure 4.:
The time to the initial and subsequent postoperative respiratory depressive episodes. PRODIGY indicates The PRediction of Opioid-induced respiratory Depression In patients monitored by capnoGraphY; RD, respiratory depression.

A secondary analysis was performed to determine if the number of RD events changed with differing PRODIGY risk category scores. In this analysis, it was found that the proportion of patients who had RD episodes and the number of episodes per patient increased with increasing PRODIGY risk category (Table 2; Figure 4).


This data analysis of continuous capnography and pulse oximetry obtained from 2 cohorts of postoperative patients on standard general care inpatient wards yielded several important observations regarding postoperative RD. First, patients rarely have isolated RD events with many having multiple events during the monitoring period. Second, the highest frequency of the initial confirmed RD events occurred in the afternoon (14:00–20:00), several hours after surgery completion. These initial RD were associated with numerous subsequent RD episodes which occurred during the same timeframe. The frequency of all confirmed RD peaked during the early morning hours. These observations were present despite that in many cases, the applications of the Capnostream monitor were delayed for several hours following general care ward admission, resulting in the possibility that many early cases of RD were undetected. This observation suggests that if continuous postoperative monitoring is to be used, the best practice would be the application of these monitors in the PACU will be used, should be initiated before discharge from the PACU. Finally, the number of RD events per patient correlated with the PRODIGY risk score.

The methodology of this post hoc analysis of a subset of data differed from the primary PRODIGY study3 methodology vis-à-vis adjudication of suspected RD events recorded by the Capnostream monitor. During the initial PRODIGY trial, the investigators charged with adjudicating potential RD episodes from the Capnostream monitor noted the number of subsequent RD events that often occurred in patients after the initial confirmed RD (personal observation of senior author). Thus, for practical considerations, data review of Capnostream alerts was limited to the occurrence of the first confirmed episode of RD, and patients were dichotomized as having ≥1 RD episode or none. In this post hoc study, all episodes of potential RD for each include patient were reviewed and adjudicated. This secondary review confirmed that patients experiencing at least 1 RD episode, rarely only had just 1, with 1 patient having more than 500 RD episodes. For patients with multiple RD events, the typical pattern observed was repetitive episodes of apnea. However, as this study was limited to only 2 of the PRODIGY study centers, it is unclear if these observed patterns of repetitive RD events can be extrapolated to other practices. Compared to other PRODIGY sites, subjects in the current study were younger, but had a higher body mass index, PRODIGY risk scores, and American Society of Anesthesiologist physical status score and underwent more extensive surgeries. However, the findings by Sun et al5 are germane to the question of generalizability. In that study, 833 postoperative patients at both a major academic hospital and a community hospital underwent continuous monitoring with pulse oximetry (the health care teams were blinded to the monitors). Hypoxemic episodes were common and prolonged with 37% of patients having Spo2 readings <90% for over an hour.5 Because monitoring technology between the current study and the Sun et al5 differed, outcomes are not completely translatable (repetitive apneas versus prolonged hypoxemia), but their similarities support the notion that RD episodes frequently are not isolated events.

The PRODIGY score is calculated from 5 clinical factors and assigns patients to low-, intermediate-, or high-risk categories for opioid-induced RD where there is a 2-fold increase in risk from low to intermediate category (odds ratio [OR] = 2.34 [95% CI, 1.72-3.19]; P < .001) and a 6-fold increase in risk from low to high category (OR = 6.07 [95% CI, 4.44-8.30]; P < .001).3 The present study demonstrated that not only is the PRODIGY score predictive of increased risk of RD (an expected finding given that these subjects were used to derive the score), but that the number of RD events per patient increases as the risk score increases. Though this study was not designed to determine if increasing number of RD episodes increases risk for serious morbidity (eg, multiple RD episodes culminating as a respiratory arrest), we speculate that this relationship may exist. This observation that patients with higher scores had more RD episodes adds credence to the notion that patients with high PRODIGY risk scores should undergo more intensive postoperative monitoring. An unknown question is if patients without a previous assessment for sleep-disordered breathing but have episodes of RD, should have later investigation for such disorders. Elleby et al12 found that 24% of patients who had isolated episodes of RD in the PACU and were subsequently evaluated by polysomnography had a 24% rate of severe obstructive sleep apnea.

The temporal frequency distribution of postoperative RD episodes, with a peak of initial episodes in the afternoon/evening, and subsequent episodes during the early morning, observed in this study provides insight into 2 previous observations regarding postoperative RD. Studies that examined postoperative naloxone administration as surrogate for severe opioid-induced RD have noted that a substantial proportion of naloxone interventions occur in the first few hours following surgery.8,10 Ramachandran et al9 observed a peak occurrence of inpatient respiratory arrests between 12:00 and 17:59, which seem to be temporally related to the peak times of PACU discharges rather than a function of circadian sleep-wake cycle.13 In the present study, the highest frequency of initial RD episodes occurred in the first few hours after surgery typically in the late afternoon and early evening, which lagged a few hours from the end of their surgeries earlier in the afternoon. Also, a large number of subsequent RD events occurred during the afternoon and evening. It is unclear from these data if these earlier episodes differed in character from those that occurred later in the early morning, and if 1 type is more likely to result in respiratory failure. This is an important question because if there are phenotypic differences in RD episodes between those that occur early versus later, it might explain why such a large proportion of severe opioid-induced respiratory complications tend to occur in the first few postsurgical hours.8–10 The practical implication of these observations is that postoperative patients who need continuous monitoring, or those determined to be at intermediate or high risk per PRODIGY score, should have these monitors applied on arrival to the postoperative ward.14

The high frequency of RD episodes observed during early morning hours supports the findings by Chung et al7 that sleep parameters as measured by polysomnography in patients with and without sleep-disordered breathing worsen postoperatively, with peak central sleep apnea episodes occurring during the first postoperative night and peak apnea-hypopnea index on the third postoperative night. Data from the current study were mostly limited to the first postoperative night, so it is unknown if there would be another period of high frequency of RD episodes the following nights.

This study had several limitations. The timing of the application of the Capnostream monitor was not mandated in the initial study, nor was the time of admission to the general care ward recorded. Thus, RD episodes may have occurred without detection before monitor application, and therefore, the time to initial RD episode may be overestimated. The median duration of monitoring was about 1 day, so we cannot comment if the observed temporal patterns would repeat on subsequent postsurgical days. Preoperative Capnostream data were not collected, so it is unknown if abnormal breathing patterns observed during sleep represented preexisting sleep-disordered breathing conditions or if those conditions worsened following surgery. Continuously collected capnography and pulse oximetry data are subject to artifact and further limited by patient compliance; thus, a degree of subjectivity is involved in adjudicating potential RD episodes. Detailed information regarding timing of opioid analgesic or other sedating medication administration was not available from existing data, so potential relationships could not be explored. Finally, because of resource constraints, it was feasible to review data from 2 of the participating institutions. Patients from these 2 centers had higher acuity than subjects from other participating centers, so it is unclear if our results would be generalizable to other practices.

In conclusion, a large proportion of the study cohort of postsurgical patients continuously monitored by bedside capnography and pulse oximetry demonstrated frequent and repetitive RD episodes. The frequency of these RD episodes increased with higher PRODIGY risk scores. Initial postoperative RD episodes occurred in the first several hours following surgery with a peak occurrence in the late afternoon and early evening and were associated with a large number of subsequent RD episodes. The peak occurrence of all RD episodes occurred during the early morning hours. This information could be useful in devising strategies for continuous postoperative respiratory monitoring.


Katherine E. Liu, PhD (Medtronic, Minneapolis, MN) provided medical writing support.


Name: C. Noelle Driver, BS.

Contribution: This author helped analyze and interpret data, draft the manuscript, and revise the manuscript.

Conflicts of Interest: None.

Name: Mariana L. Laporta, MD.

Contribution: This author helped analyze and interpret data, draft the manuscript, and revise the manuscript.

Conflicts of Interest: None.

Name: Sergio D. Bergese, MD.

Contribution: This author helped acquire and interpret data and revise the manuscript.

Conflicts of Interest: S. D. Bergese reports financial support to the Investigator or Investigator’s Institution to fund the Medtronic-sponsored trial.

Name: Richard D. Urman, MD.

Contribution: This author helped acquire and interpret data and revise the manuscript.

Conflicts of Interest: R. D. Urman reports financial support to the Investigator or Investigator’s Institution to fund the Medtronic-sponsored trial. In addition, R. D. Urman reports a grant and personal fees from Merck and personal fees from 3M, Posimir, Medtronic, AcelRx, Heron, Acacia, and Takeda.

Name: Fabio Di Piazza, MS.

Contribution: This author helped analyze data and revise the manuscript.

Conflicts of Interest: F. Di Piazza reports employment with Medtronic.

Name: Frank J. Overdyk, MD.

Contribution: This author helped acquire and interpret data and revise the manuscript.

Conflicts of Interest: F. J. Overdyk reports financial support to the Investigator or Investigator’s Institution to fund the Medtronic-sponsored trial.

Name: Juraj Sprung, MD, PhD.

Contribution: This author helped interpret data and revise the manuscript.

Conflicts of Interest: None.

Name: Toby N. Weingarten, MD.

Contribution: This author helped design the analysis, analyze and interpret data, and revise the manuscript.

Conflicts of Interest: T. N. Weingarten reports financial support to the Investigator or Investigator’s Institution to fund the Medtronic-sponsored trial and a grant from Merck and nonfinancial support from Respiratory Motion.

This manuscript was handled by: Jean-Francois Pittet, MD.


    1. Oderda GM, Gan TJ, Johnson BH, Robinson SB. Effect of opioid-related adverse events on outcomes in selected surgical patients. J Pain Palliat Care Pharmacother. 2013;27:62–70.
    2. Dahan A, Aarts L, Smith TW. Incidence, reversal, and prevention of opioid-induced respiratory depression. Anesthesiology. 2010;112:226–238.
    3. Khanna AK, Bergese SD, Jungquist CR, et al.; PRediction of Opioid-induced respiratory Depression In patients monitored by capnoGraphY (PRODIGY) Group Collaborators. Prediction of opioid-induced respiratory depression on inpatient wards using continuous capnography and oximetry: an international prospective, observational trial. Anesth Analg. 2020;131:1012–1024.
    4. Overdyk FJ, Carter R, Maddox RR, Callura J, Herrin AE, Henriquez C. Continuous oximetry/capnometry monitoring reveals frequent desaturation and bradypnea during patient-controlled analgesia. Anesth Analg. 2007;105:412–418.
    5. Sun Z, Sessler DI, Dalton JE, et al. Postoperative hypoxemia is common and persistent: a prospective blinded observational study. Anesth Analg. 2015;121:709–715.
    6. Benumof JL. Mismanagement of obstructive sleep apnea may result in finding these patients dead in bed. Can J Anaesth. 2016;63:3–7.
    7. Chung F, Liao P, Yegneswaran B, Shapiro CM, Kang W. Postoperative changes in sleep-disordered breathing and sleep architecture in patients with obstructive sleep apnea. Anesthesiology. 2014;120:287–298.
    8. Deljou A, Hedrick SJ, Portner ER, et al. Pattern of perioperative gabapentinoid use and risk for postoperative naloxone administration. Br J Anaesth. 2018;120:798–806.
    9. Ramachandran SK, Haider N, Saran KA, et al. Life-threatening critical respiratory events: a retrospective study of postoperative patients found unresponsive during analgesic therapy. J Clin Anesth. 2011;23:207–213.
    10. Weingarten TN, Herasevich V, McGlinch MC, et al. Predictors of delayed postoperative respiratory depression assessed from naloxone administration. Anesth Analg. 2015;121:422–429.
    11. Khanna AK, Overdyk FJ, Greening C, Di Stefano P, Buhre WF. Respiratory depression in low acuity hospital settings-seeking answers from the PRODIGY trial. J Crit Care. 2018;47:80–87.
    12. Elleby LM, Sprung J, Weingarten TN. Postoperative respiratory depression may be related to undiagnosed sleep-disordered breathing or obstructive sleep apnea. Anesth Analg. 2019;129:e104–e105.
    13. Weingarten TN, Warner LL, Sprung J. Timing of postoperative respiratory emergencies: when do they really occur? Curr Opin Anaesthesiol. 2017;30:156–162.
    14. Ayad S, Khanna AK, Iqbal SU, Singla N. Characterisation and monitoring of postoperative respiratory depression: current approaches and future considerations. Br J Anaesth. 2019;123:378–391.

    Supplemental Digital Content

    Copyright © 2021 The Author(s). Published by Wolters Kluwer Health, Inc. on behalf of the International Anesthesia Research Society.