PRODIGY score predicts respiratory depression in the post-anesthesia care unit: A post-hoc analysis

,


INTRODUCTION
The initial phase of anesthesia recovery is complex, where vital organ systems regain normal function from the effects of general anesthesia and surgery.Typically, this process occurs in the postanesthesia care unit (PACU), a specialized unit where patients receive intensive level care from specialized nurses under the supervision of an anesthesiologist [1].Emerging evidence has found strong correlations between respiratory depression episodes (RDE, eg episodes of apnea) occurring during the PACU admission and subsequent development of potentially life-threatening episodes of respiratory failure on general care postoperative wards [2].Preoperative assessment of patient risk for PACU RDEs could be of clinical benefit to the anesthesiologist in developing an anesthetic plan to mitigate this risk; however, no such scoring tools have been assessed.
The PRediction of Opioid-induced respiratory Depression In patients monitored by capnoGraphY (PRODIGY) trial was a large, prospective multinational trial which utilized continuous capnography and pulse oximetry to monitor patients on general care wards in order to develop a prediction model for RDE risk [3].In that study, RDE was defined as respiratory rate ≤5 breaths/min, oxyhemoglobin saturation ≤85%, or end-tidal carbon dioxide ≤15 or ≥60 mm Hg for ≥3 minutes; apnea episode >30 seconds; or any respiratory opioid-related adverse episode, such as opioid reversal, respiratory failure, or cardiopulmonary arrest.Using these data, a multivariable respiratory depression prediction model (PRODIGY score) was developed using 5 independent variables (age, sex, history of sleep disordered breathing, chronic heart failure, and opioid naïvety, Supplementary Table 1).The PRODIGY score can be used to categorize patients as low, intermediate, or high-risk for RDEs on the general care wards [3].
Previously, Chung et al [4] utilized continuous capnography and pulse oximetry on adult postsurgical patients to characterize RDEs in the PACU, following general anesthesia.Because this dataset utilized the same technology (capnography and pulse oximetry) to diagnose RDE as used in the PRODIGY trial, it provides a unique opportunity to assess if the PRODIGY score can be used to assess risk for PACU RDEs.In the current study, we conducted a post-hoc analysis of a subset of the subjects enrolled in Chung et al [4] to explore if there is a rationale to consider if the PRODIGY score is associated with RDEs in the PACU.

MATERIALS AND METHODS
We performed a post-hoc analysis of the Chung et al [4] dataset to assess whether the PRODIGY score is associated with RDEs in the PACU.

Participants
The original trial was designed to include up to 250 patients across two sites.This post-hoc analysis was performed using patient data from the United States trial site only, including the 87 patients who completed the trial and had complete (≥90%) monitoring data at that site [4].As in the original trial, patients with >10% of continuous monitoring data missing were excluded from the analysis cohort, as imputation of the missing monitoring data was not feasible.Second trial site data was not available for post-hoc evaluation due to data privacy restrictions.

Trial design and objective
The trial was designed to quantify the occurrence of respiratory episodes in patients in the PACU.
The anesthetic management was under the direction of the attending anesthesiologist and included maintenance with volatile anesthetic and analgesia with fentanyl.Neuromuscular blockade, if indicated, was reversed at the conclusion of the procedure, and midazolam administered preoperatively if needed for anxiolysis.After being transferred to the PACU, patients were monitored using standard of care monitoring equipment, along with blinded capnography and pulse oximetry (Capnostream™ 20p monitor, Nellcor™ Max A pulse oximeter, and Microstream™ Smart CapnoLine™ Plus O2 sampling line [Medtronic, Inc., Boulder, CO]).All enrolled patients were required to undergo capnography monitoring for at least 45 minutes.
Monitoring parameters were pre-defined to detect RDEs which would generate a Level I alerts (requiring immediate physician attention, eg administration of naloxone, application of noninvasive positive pressure ventilation) or Level II alerts (requiring nursing attention, eg increasing supplemental oxygen, repositioning patient), (specific definitions of Level I and II alerts are provided in Table 2) [4].
The five clinical variables required to calculate the PRODIGY score (age, sex, history of heart failure, history of obstructive sleep apnea (OSA), and opioid naïvety) were available for all subjects.Each variable is assigned points which are summated to calculate the PRODIGY score (Supplemental Table 1) with a score of <8 being low risk, ≥8 and <15 intermediate risk; and ≥ 15 high risk for RDE on general care wards [3].

Outcomes measures
The objective of the original PACU trial was to quantify the occurrence of respiratory episodes, including apnea, hypoxemia, hypercapnia, hypocapnia, tachypnea, bradypnea, tachycardia, and bradycardia in surgical patients.Each anticipated respiratory episode was defined a priori, including alert thresholds appropriate for nurses (Level II) and physicians (Level I) [4].The objective of this post-hoc analysis was to assess whether the PRODIGY score is associated with the incidence of respiratory episodes in the PACU.

Ethical statement
This study was conducted in accordance with the Declaration of Helsinki and all local regulatory requirements.The protocol was approved by the institutional review board of all participating sites (Toronto Western Hospital and University of Arizona Medical Center).The original trial was registered with www.clinicaltrials.govunder NCT02707003 (Principal Investigators: Peter R. Lichtenthal, Frances F. Chung; Registered March 11, 2016).All subjects provided written informed consent.This manuscript adheres to the Strengthening the Reporting of OBservational studies in Epidemiology (STROBE) guidelines.

Statistical analysis
Analyses were performed using SAS® Version 9.4 (SAS Institute Inc., Cary, NC) and data were summarized by descriptive statistics (mean with standard deviation or median with interquartile range [IQR] for continuous variables) and frequencies and percentages (for categorical variables).
ANOVA and Chi Square were used to compare baseline characteristics across PRODIGY groups.
Because the outcome of interest, PACU RDEs, was skewed, analysis across PRODIGY groups was performed with the Kruskal-Wallis test.In order to take into account varying lengths of PACU stay, Poisson regression was used to assess whether the rate of respiratory depression episodes (RDE) differed between PRODIGY groups.For this analysis, the RDE count was the dependent variable, PRODIGY group was the independent variable, and the log of PACU stay duration was included as an offset to account for patients having different PACU stay durations.Results from this analysis are summarized by presenting the rate ratio estimate with 95% confidence interval for the pairwise PRODIGY group comparisons.

Participants
Of the 125 patients assessed for eligibility and enrolled in the single site in the United States, 87 patients completed the trial (Figure 1).Eleven patients did not meet the trial inclusion and exclusion criteria, and 26 were withdrawn due to technical issues (study coordinator or monitoring equipment unavailable at time of PACU admission) or invalid device data (<90% of continuous monitoring data).All 87 patients with valid device data were included in this post-hoc analysis.
Patient and procedural characteristics are shown in Table 1.Even though the trial was open to ASA IV patients, none were recruited from the center included in this subset.

Respiratory episodes in the PACU
During anesthesia recovery, 3,580 respiratory episodes occurred during blinded capnography monitoring, including 3,050 Level II alerts and 530 Level I alerts.The most common alert was the Level II apnea alert (85% patients [n = 74/87], with a median [IQR] 13  alerts during the PACU stay) (Table 2).Bradypnea, hypocapnia, tachypnea, and hypoxemia Level II alerts were also common.Hypoxemia and bradypnea were the most common Level I alerts (Table 2).

Association of PRODIGY score and respiratory episodes
Twenty-six (29.9%) patients had low, 29 (33.3%)intermediate, and 32 (36.8%) high PRODIGY scores (Table 3).The total number and duration of alerts were highest in the high PRODIGY risk group patients (P = 0.035 and P = 0.042, respectively) (Table 3).Among patients who had a Level I alert, patients in the high PRODIGY risk group had significantly more episodes (P = 0.046, Figure 2A).Level II alerts were also most common in patients in the high PRODIGY risk group (P = 0.057, Figure 2B).From Poisson regression, the rate of RDE for patients in the high PRODIGY risk group was significantly higher than that for patients in the low (rate ratio = 2.25, 9 95% CI = 2.07 to 2.45, P<0.001) and intermediate (rate ratio = 2.01, 95% CI = 1.86 to 2.18, P<0.001) risk groups, and the rate of RDE in patients in the intermediate risk group was significantly higher than patients within the low risk group (rate ratio = 1.12, 95% CI 1.02, 1.23, P=0.022).

DISCUSSION
This study found that the number of RDEs in the PACU increase with higher PRODIGY scores.
Importantly, the rate of Level I episodes, the more serious episodes that would prompt physician intervention, were significantly greater in patients in the high PRODIGY group.These findings suggest that PRODIGY score may have utility in assessing risk of postoperative respiratory depression (PRD) during anesthesia recovery.Our results are similar to a post-hoc analysis of nursing-diagnosed RDEs in the PACU following laparoscopic surgery, where high versus low PRODIGY groups were associated with increased risk of postoperative respiratory depression (odds ratio 1.46) [5].
Failure to identify and treat PRD can lead to cardiopulmonary arrest, anoxic brain injury, and death [6,7].While continuous monitoring of all postoperative patients has been shown to improve patient outcomes, this may not be feasible in all practices or in low-resource/low-staff settings [8,9].Surgical patients who have a positive assessment screen for OSA and experience respiratory depression in the PACU have higher rates of postoperative pulmonary complications compared to other surgical patients [10].Evaluation of PRODIGY scores to assess the risk of PRDperhaps in conjunction with STOP-BANG [11] scores in patients to assess OSAin the PACU can guide postoperative monitoring decisions and help optimize the perioperative management of surgical patients based on individual needs.This could include continuous monitoring in the PACU and surgical unit or in some cases, may contribute to the decision to transfer a high-risk patient to a higher acuity unit.
Other studies have found that patients with respiratory depression in the PACU have five-fold increased risk of requiring naloxone to treat severe opioid toxicity on the general care ward [1,12].Continuous measurements of minute volume found that patients with depressed ventilation in the PACU continued to have depressed ventilation for hours once discharged to general care wards [13].This relationship between RDEs in the PACU and the ward suggests that PRODIGY score may have utility to identify patients who are at increased risk to develop respiratory depression in the PACU.Postoperative patients in the high PRODIGY group should undergo closer nursing observation in the PACU for signs of respiratory depression, and those patients who do have an RDE in the PACU should be carefully evaluated prior to discharge to general care wards to determine if enhanced postoperative monitoring and care are indicated.
Based on prior studies, enhanced monitoring may be most beneficial in the first postoperative day.
Recent studies have found that RDEs begin to occur shortly after PACU discharge to general care wards [13,14].Studies examining the use of postoperative naloxone administration to reverse severe opioid toxicity found that most naloxone was given in the afternoon and evening following surgery [2].
This study has several limitations.First, this was a sub-analysis of a previous pilot study, as data from only one site was available for analysis.Thus, this had reduced the statistical power to determine if PRODIGY score would correlate with postoperative outcomes [4].Given the limited sample size and single setting of the current study, additional studies should be performed to confirm generalizability of these findings.Second, capnography and pulse oximetry monitoring were confined to the PACU only, thus we can only speculate that patients who had respiratory depression in the PACU continued to do so on the wards.Thirdly, ASA I and IV patients were not included in this analysis.Inclusion or subset analysis of ASA IV patients may further lend clues into risk factors that may lead to increased risk of RDEs.Lastly, because the anesthetic care was left to the discretion of the attending anesthesiologist, we cannot account if perioperative management was altered based on perception of risk for PACU RDEs.

CONCLUSION
A high PRODIGY score, designed to detect RDEs on general care wards, maybe associated with RDE occurrence in the PACU.These observations suggest the potential utility of the PRODIGY score in assessing the risk for early RDEs occurring during anesthesia recovery, but further study is required to determine if these findings can be generalized to other populations.disorder, asthma, or other pulmonary/respiratory disease.ASA: American Society of Anesthesiologists; PACU: Post anesthesia care unit; PRODIGY: PRediction of Opioid-induced respiratory Depression In patients monitored by capnoGraphY; SD: Standard deviation.

FIGURE 2 .
FIGURE 2. Assessment of Level I (A) and Level II (B) alerts, by PRODIGY Risk group (high, intermediate, or low).PRODIGY: PRediction of Opioid-induced respiratory Depression In patients monitored by capnoGraphY.

TABLE 2 . Types of respiratory adverse episodes and frequency of respiratory episodes in the postanesthesia care unit.
Level I and Level II alerts are categories of respiratory depression episodes with Level I defined as requiring immediate physician attention and Level II defined as requiring nursing attention.bpm: Breaths per minute; IQR: Interquartile range.