Key action points in the management of postoperative respiratory depression include the following:
The incidence of postoperative respiratory depression varies according to the defining criteria employed (eg, use of pharmacologic reversal with agents such as naloxone, hypoventilation, hypercapnia, and hypoxemia, in varying degrees)
Besides "too much anesthesia," factors known to be commonly responsible for respiratory compromise include those specific to the type and duration of surgery and anesthesia, as well as consequent pain, splinting, abdominal wall binding, and abdominal distention; other factors are patient-specific (eg, age, body habitus, and preexisting pulmonary disease)
The effects of residual muscle relaxants may play a role in the occurrence of immediate postoperative respiratory events in the postanesthesia recovery unit (PACU)
Hypoxemia beyond the PACU and on the postoperative surgical floor is common, persistent, and largely undetected
Neither the STOP-Bang score (as a surrogate measure of the severity of obstructive sleep apnea [OSA]) nor the use of long-acting opioids in patient-controlled analgesia (PCA) systems seems to predict the development of hypoxemia in the postoperative period
The unpredictability of postoperative hypoxemia demands that respiratory monitoring be universally implemented on the nursing floor
This respiratory monitoring should be all-inclusive, continuous, and automated, and it should address multiple parameters, including oxygen saturation, ventilation, and cardiac parameters (at a minimum, heart rate [HR], as well as blood pressure [BP] if possible)
Intraoperative mortality is rare. Unfortunately, the same cannot be said of postoperative mortality. Approximately 2% of all postoperative inpatients die within 1 month.  If the 30-day period after a surgical procedure were considered a disease, it would be the third leading cause of death in the United States. 
Cardiac complications are by far the most frequent cause of 30-day postoperative mortality. However, respiratory depression in the postoperative period is also common. In its 2015 report, the Agency for Healthcare Research and Quality (AHRQ) rated postoperative respiratory failure as the fourth most common patient safety event and the second most common if obstetric indicators were excluded. 
A degree of ambiguity continues to surround the definition of respiratory depression. A combination of naloxone use, hypoventilation, hypercarbia, and hypoxemia has been used to define this physiologic aberration, but different sources have specified different durations and thresholds for these factors. Consequently, the reported incidence has ranged from 0.3% to 17%, depending on the specific defining criteria employed. 
Immediate postoperative respiratory depression is seen in the PACU. The most common causes in this setting are usually a direct consequence of ''too much anesthesia." In varying combinations, these causes may include excessive opioid use, inadequate reversal of neuromuscular blockade, or prolonged persistence of these agents in the system as a result of liver or renal disease. In addition, the lingering effects of inhalational agents and long-acting benzodiazepines may contribute to the development of respiratory depression in the PACU.
The other common group of causes of postoperative respiratory depression includes those specific to the type and duration of surgery and anesthesia and consequent pain, splinting, abdominal wall binding, and abdominal distention. Finally, a number of patient-related factors may play a role, including age, body habitus, and preexisting pulmonary disease.
Whereas postoperative patients commonly find themselves in the PACU and the intensive care unit (ICU), these areas are intensively monitored as a rule. It goes without saying that the destination for these patients, once they have left the PACU and before their condition has deteriorated sufficiently to warrant ICU admission, is the regular nursing floor or ward. This is currently monitored via nursing checks on vital signs every 4-6 hours. Generally, the regular ward is a place of comfort for clinically stable patients.
In a prospective blinded observational study by Sun et al that was aimed at quantifying hypoxemia on surgical wards in patients who had undergone noncardiac surgery,  continuous pulse oximetry was used to monitor postoperative oxygen saturation for as long as 48 hours after the procedure.  This continuous monitoring started once the patient left the PACU or the ICU and reached the regular nursing floor. Bedside care providers were blinded to the findings from oximetry. The nurses continued their routine checks on vital signs every 4 hours per protocol.
The results told a compelling story: Postoperative hypoxemia was common, serious, and prolonged (see Table 1 below).  For example, 20% of patients demonstrated an average of 10 minutes per hour of oxygen saturation below 90% over their entire hospital stay. It is sobering—and somewhat shocking—that 90% of serious hypoxemic episodes (oxygen saturation <90% for ≥1 hour) were completely missed by nurses and physicians conducting routine vital sign monitoring at 4-hour intervals.
Table 1. Incidence of Postoperative Hypoxemia on Regular Ward (Open Table in a new window)
|Blinded SpO2||Duration (over 48 hr of monitoring)||% of patients monitored|
|SpO2 = oxygen saturation measured by pulse oximetry.|
Patients continue to decompensate on the regular nursing floor, resulting in emergency medical team activation and transfer to higher levels of care. The problem is so serious that it has led to the formation of a "dead-in-bed" registry (http://www.newsnet5.com/longform/dead-in-bed-a-deadly-hospital-secret). Postoperative hypoxemia not only is alarmingly common but also remains frustratingly difficult to anticipate.
Patients with OSA typically have cyclical oxygen desaturations that are characteristic of the disease. Surprisingly, a large cohort analysis determined that STOP-Bang (Snoring, Tiredness, Observed apnea, high blood Pressure–Body mass index, Age, Neck circumference, Gender) scores—a validated measure of OSA risk—were not associated with the degree of postoperative oxygen desaturation. 
Narcotic use—or, rather, overuse—is often regarded as the causative factor behind emergency transfers from the regular floor to the ICU. A group that included the author examined the association between postoperative hypoxemia and the type of narcotic used (ie, long-acting [morphine or hydromorphone] or short-acting [fentanyl]) in PCA systems and found that the use of short-acting opioids did not reduce the risk of hypoxemia. 
Thus, available robust evidence appears to suggest that reliable predictions cannot be made regarding which postoperative inpatients will become hypoxemic, how severe the hypoxemia will be, or what pattern the desaturation will follow.
The ARISCAT (Assess Respiratory rIsk in Surgical patients in CATalonia) risk index (including age, low preoperative oxygen saturation on pulse oximetry in air, respiratory infection in the past month, preoperative anemia, upper abdominal or intrathoracic surgical incision, duration of surgery ≥2 hours, and emergency procedure) has been validated as a useful predictor of the risk of postoperative pulmonary complications (PPCs). [8, 9]
However, the patients in these studies were at a baseline level of risk, and complications that involved specific pulmonary pathology were identified as outcomes. Indeed, the authors did not evaluate the risk of respiratory arrest and unforeseen respiratory compromise (desaturation, hypoventilation, or a combination thereof) in healthy patients on the regular nursing floor. [8, 9]
Addressing the problem
Given that postoperative respiratory depression remains a common, unpredictable, and frequently unforeseen problem, the obvious question that arises is, How can this problem be solved?
A number of catastrophic respiratory events may be avoided by employing continuous automated cardiorespiratory monitoring. How, what, and whom to monitor are questions that remain to be answered.  In a large tertiary care medical center patient cohort, manually recorded oxygen saturation values were, on average, 6.5% higher than those recorded by automated systems. 
Although continuous pulse oximetry on the regular ward certainly prevents ICU transfers and decreases rescue events, it is by no means the be-all and end-all of respiratory monitoring.  The American Society of Anesthesiologists (ASA) recommends continuous monitoring of oxygenation and ventilation in patients with neuraxial blocks and extended monitoring in patients with OSA. [13, 14]
In addition, the heart is never too far away from the lungs, and it is fair to say that respiratory events do not occur in isolation. In a struggling patient, tachycardia or hypoxemia and hypoventilation may occur early on, either separately or collectively, and they often progress to hypotension, which in itself is strongly associated with myocardial injury and death. [15, 16] As a corollary, it is well established that vital signs deteriorate 6-12 hours before cardiac and respiratory arrests occur; this is the rationale for having hospital rapid-response teams, which undoubtedly save lives in these situations. [17, 18, 19]
Accordingly, it is reasonable to conclude that a combination of oxygenation, ventilation, and a minimum hemodynamic parameter should be monitored in every patient in a continuous automated fashion.
Many more dimensions remain to be explored. For instance, optimal handling of monitors and the continuous stream of generated numbers, reduction of "noise" and false alarms, institution of centralized alarm monitoring systems, use of early effective warning scores, and prevention of alarm fatigue are certainly important educational and workflow pieces for colleagues on the regular nursing floors.
It may seem that an easy solution to the problem would be simply to leave every borderline patient in the PACU or to transfer everyone to the ICU, but this would undoubtedly be less than optimal care. There is a "gray area" beyond the confines of the PACU and before the doors of the ICU, and patients in this area deserve better monitoring to prevent postoperative respiratory depression.
In a 2015 closed claims analysis that examined postoperative opioid-induced respiratory depression, at least 77% of the patients died or sustained severe brain damage.  Only 9% had abnormal STOP-Bang scores. It is crucial to note that 97% of these adverse events were deemed to have been preventable with better monitoring and response and that 42% of the episodes occurred within 2 hours of the most recent nursing check. 
All of this evidence points to the dangers of unpredictable and unforeseen respiratory depression in the unmonitored environment of the regular nursing floor and underscores the importance of continuous and efficient multiparameter monitoring of patients in these areas.
With regard to effective prevention of postoperative respiratory disasters, there are still more questions than answers. When the offending agent is clear, management may be straightforward, following the principles of ventilatory support and pharmacologic reversal of the offending agent. However, this seems not to be the usual cause of deterioration when patients leave the PACU, and management must often be more complex.
The PRODIGY (PRediction of Opioid-induced Respiratory Depression In Patients Monitored by capnoGraphY) trial may help answer some of those questions and enable better prediction, monitoring, and treatment of respiratory depression. In any case, we anesthesiologists owe it to our patients to carry out a safe, secure, and reliably monitored care plan while they recover on medical and surgical wards.
Multimodal analgesia, as much as is allowable
Improved monitoring and feedback response systems for care providers and emergency responders on the regular floors
Continuous pulse oximetry
Early identification of perceived risk factors, with the knowledge that there is a need to develop a validated risk prediction score based on deranged respiratory vital signs
Education and orientation of care providers
Continuous supplemental oxygen for at least those patients perceived to be at high risk (eg, because of sleep apnea or use of intravenous [IV] narcotics for PCA)
Case Example 1
A 53-year-old man is scheduled to undergo a laparoscopic cholecystectomy. His medical history is notable for the following:
Hypertension - Controlled on daily metoprolol
Hyperlipidemia - Addressed with a statin regimen
Hypothyroidism - The most recent thyroid-stimulating hormone (TSH) level (obtained 1 week previously) was normal, and the patient is taking thyroxine 75 μg/day
Chronic renal failure - The patient is dialysis-dependent and was last dialyzed yesterday; morning laboratory tests were done, and the potassium concentration is 5.1 mEq/L
Anesthesia is induced with propofol and fentanyl, and rocuronium is used for intubation of the trachea. The duration of the operation is brief (50 minutes) and its course unremarkable. Anesthetic maintenance is with isoflurane in combination with air and oxygen. At the end of the procedure, the patient has two out of four twitches. Reversal is accomplished with neostigmine and glycopyrrolate in weight-appropriate doses. Subsequently, the patient generates good tidal volumes and exhibits criteria for return of muscle strength. He is then successfully extubated.
Twenty minutes later, the staff anesthesiologist receives an emergency page to the bedside from the PACU. The patient is now gasping, using all accessory muscles of breathing and unable to move much air. A bedside examination shows poor hand grip strength and lack of adequate muscle strength. In addition, he has lost the ability to generate a good cough. Vital signs are as follows:
HR, 130 beats/min
Respiratory rate (RR), 40 breaths/min
BP, 180/100 mm Hg
SpO 2, 80% on a 100% nonrebreather mask
A decision was made to initiate bag-valve-mask ventilation. Every small respiratory effort the patient made was supported with a gentle breath from the bag-mask assembly. Meanwhile, IV sugammadex was administered and brought about immediate resolution of symptoms. After a chest radiograph was rechecked and return of muscle strength noted, the patient was placed on 2 L of oxygen via nasal cannula (NC) and made an uneventful recovery in the PACU.
This vignette illustrates an instance of immediate postoperative respiratory depression. The likely cause was residual muscle relaxant. Because of the patient's impaired renal function, the intubating dose of rocuronium may not have been totally removed from his system. In addition, there appears to have been some residual paralysis when the reversal agent was administered. Thus, it is probable that the neostigmine transiently occupied the nerve muscle end-plate receptors, only to be overwhelmed by the residual rocuronium (a phenomenon known as delayed recurarization).
After ventilatory support was instituted, reversal was achieved by administering sugammadex. This agent is a selective relaxant-binding drug that encapsulates the nondepolarizing aminosteroid muscle relaxants rocuronium and vecuronium, reversing and preventing their action.
Case Example 2
A 20-year-old woman is scheduled to undergo closure of an ileostomy. She has Crohn disease and has undergone at least 10 major abdominal surgical procedures in the past 10 years. At home, she is heavily dependent on pain medications, both narcotic and nonnarcotic. In addition, she is taking gabapentin, amitriptyline, and diazepam for anxiety and a mood disorder.
Anticipating that management of the patient's postoperative pain will prove difficult, the anesthesiologist places a lumbar epidural catheter preoperatively and institutes epidural PCA with morphine. A ketamine infusion is also started intraoperatively. The operation takes 3 hours but is largely uneventful. Reversal is achieved, and the patient is extubated.
In the PACU, the patient is receiving morphine via epidural PCA, as well as continued IV ketamine infusion, but is still complaining of pain that she rates as 10/10. She is then started on IV PCA with fentanyl and is given additional doses of fentanyl by the clinician. After spending 4 hours in the PACU, she meets PACU discharge criteria and is sent to the regular floor.
Later that evening, the patient complains of anxiety and is given two doses of IV diazepam, another clinician dose of IV fentanyl, and her regular doses of gabapentin and amitriptyline. At 10:00 PM that night, a medical emergency team is called to the bedside because the patient's nurse has found her breathing shallowly and blue while on a floor bed.
The medical emergency team, led by a critical care anesthesiologist, was quick to realize that this was an instance of narcotic overdose secondary to the cumulative effect of narcotics administered via different routes and topped up with additional IV boluses and nonnarcotic sedative drugs in a relatively unmonitored environment.
As a first step, the patient's ventilation was supported with a bag-valve-mask system with supplemental oxygen. Next, IV naloxone was administered, which immediately resolved her symptoms. As a further step, the epidural was stopped, and an opioid-free epidural solution was employed instead. The IV ketamine drip was rapidly tapered down. The pain regimen was also strengthened with the use of IV acetaminophen.
Because naloxone is a short-acting medication in comparison with the narcotics that accumulated in her system, the patient was transferred to a monitored environment in an ICU bed. Her diazepam prescription was written so as to specify that doses may not be repeated within a 24-hour period. The patient then completed an uneventful and pain-free recovery in the hospital.
Case Example 3
An otherwise healthy 75-year-old man is scheduled to undergo a laparoscopic ventral hernia repair. He is remarkably healthy, exercises regularly, and has borderline diabetes that is being controlled via dietary changes and lifestyle modifications. The patient's wife says that he snores at night, but his preoperative STOP-Bang score is about 3 (low risk). The anesthesia and the surgical procedure are totally uneventful. A conscious decision is also made to use only short-acting narcotics (fentanyl) in combination with nonnarcotic medications on the floor.
On postoperative day 2, the nurse is attending to another individual who is sharing the patient's room when she hears the patient gasping and struggling to catch his breath. She checks his vital signs, noting an SpO2 of 89% on room air and an RR of 9 breaths/min. She puts the patient on supplemental oxygen, after which his SpO2 rises to 93% and he appears alert. The nurse then returns her attention to the other individual. Later that night, a code blue is called on the patient for cardiorespiratory arrest.
After one round of chest compressions and immediate ventilation with endotracheal intubation, the patient experienced an immediate return of spontaneous circulation. He was then transferred to the ICU and remained ventilated overnight. Subsequently, he was extubated and made an uneventful recovery.
This scenario is complicated, in that there was very little in the patient's history or preoperative examination that would have put him at increased risk for postoperative respiratory depression. His OSA screening score was borderline, and he was receiving only short-acting narcotics for pain control. Here is a clear example of unforeseen hypoxemia and hypoventilation that happened outside the timeframe of the routine q4hr nursing checks and was detected only because the nurse was nearby attending to another patient and happened to overhear his difficulties.
The nurse's intervention to put this patient on supplemental oxygen was appropriate in that it improved his oxygenation, but it did nothing for his lack of adequate ventilation. Furthermore, he had been having periods of hypoventilation and hypoxemia that were not being picked up. Ultimately, the patient experienced respiratory arrest, which may have been preceded by a cardiac event secondary to the chronic hypoxemia and hypoventilation. Fortunately, he responded to immediate intervention.
This scenario is a vivid illustration of the unpredictability of postoperative respiratory depression and makes a strong case for the need for continuous monitoring that includes cardiac and multiple respiratory parameters (eg, SpO2 and rate and depth of ventilation).