The OSA discharge rule is real, and its timing is undefined

Why it matters

Obstructive sleep apnea is the PACU diagnosis where the resident most often asks for a number, how long, and where the honest answer is that the guideline does not have one. The 2014 ASA practice guideline states, verbatim, that the literature is insufficient to offer guidance regarding the appropriate time for discharge of patients at increased perioperative risk from OSA from the surgical facility gross-2014. The recommendation that follows is nonetheless firm: such patients should not be discharged from the recovery area to an unmonitored setting, home or an unmonitored hospital bed, until they are no longer at risk of postoperative respiratory depression gross-2014.

The consequence of getting this wrong is either to invent a number the source does not contain, and teach it as if it were evidence, or to treat the absence of a number as the absence of an obligation. Both are errors. The obligation is real and the timing is undefined, and residents should be taught that shape.

Mechanism

Obstructive sleep apnea is a disorder of upper airway collapsibility during sleep. Anesthetic agents, opioids and sedatives reduce the tone of the pharyngeal dilator muscles and blunt the arousal response that would normally terminate an obstruction. Residual drug effect, the supine position and the rebound of REM sleep in the days after surgery all extend the window in which obstruction and hypoventilation can occur unobserved. The mechanism is not in dispute. What is in dispute, or rather unmeasured, is how long the window stays open in a given patient, and that is precisely what the guideline says the literature cannot tell you.

Evidence

The ASA guideline’s evidentiary statement is the sentence the day is built on: the literature is insufficient to offer guidance regarding the appropriate time for discharge gross-2014. The recommendation that follows does not carry an evidence grade. It carries an opinion qualifier: the consultants and ASA members strongly agree that patients at increased perioperative risk from OSA should not be discharged from the recovery area to an unmonitored setting until they are no longer at risk of postoperative respiratory depression gross-2014.

Two things to teach from that pair. First, the strength language attaches to agreement among consultants, not to evidence; strongly agree is a survey result gross-2014. Second, the recommendation is circular as an operational rule: it says do not discharge until the patient is no longer at risk, while the preceding sentence says the literature cannot tell you when that is gross-2014.

The 2016 Society of Anesthesia and Sleep Medicine guideline hedges in a different register. It records explicitly that few randomized studies exist in this field and that most recommendations were developed by experts through consensus processes with evidence grading, and it carries the disclaimer that the guideline may not be appropriate for all clinical situations chung-2016-sasm. That self-description is itself a teaching point, and a useful contrast to the ASA document’s opinion qualifier.

What this does not show

Neither guideline offers a duration, a monitoring modality or a discharge criterion that has been tested against an outcome. The evidence categories and levels behind the ASA guideline’s monitoring and discharge-timing recommendations have not yet been extracted from the held document, so this page cannot say which of its statements rest on which category of evidence gross-2014.

The two guidelines do not disagree; they hedge differently. The ASA document states an obligation with an opinion qualifier, and the SASM document states the limits of its own evidence base gross-2014 chung-2016-sasm. Neither is a substitute for a monitoring protocol, and neither says that any particular protocol works.

The screening instrument has its derivation. STOP asks four yes-or-no questions, Snoring, Tiredness, Observed apnea and blood Pressure; in surgical patients its sensitivity for an apnea-hypopnea index above 5, 15 and 30 was 65.6%, 74.3% and 79.5%, rising to 83.6%, 92.9% and 100% when body mass index, age, neck circumference and gender were added to make STOP-Bang chung-2008-stop. The four Bang thresholds were read from the full text on 2026-09-06 and are, verbatim, a body mass index greater than 35 kg/m2, age older than 50 years, male gender, and a neck circumference greater than 40 cm chung-2008-stop. The cut-off stated in that paper is not the STOP-Bang cut-off residents are taught: it is for STOP alone, where answering yes to two or more questions gave the best combination of sensitivity and specificity chung-2008-stop. What the score means at the top end is measured: among surgical patients with polysomnography, a STOP-Bang score of 5 carried an odds ratio of 4.8 for moderate-to-severe and 10.4 for severe sleep apnea, a score of 6 carried 6.3 and 11.6, and a score of 7 to 8 carried 6.9 and 14.9, with the predicted probability of moderate-to-severe disease rising from 0.36 at a score of 3 to 0.60 at 7 to 8; a score of 5 to 8 identifies a high probability of moderate-to-severe apnea chung-2012-stopbang. None of that supplies a monitoring duration, which is the number the guideline says does not exist.

Two figures are quoted against this day often enough to need answering here, a median of about 3 hours of extra monitoring and a median of about 7 hours after the last episode of airway obstruction or hypoxemia on room air. They are not in the current guideline. Its full text was searched for each of those phrases on 2026-09-06 and both are absent; the document carries the discharge obligation and the room-air observation and no duration at all gross-2014. The figures belong to the superseded 2006 original, and that document has now been read: the consultants indicated that patients with obstructive sleep apnea should be monitored for a median of 3 h longer than their counterparts without it before discharge from the facility, and that monitoring should continue for a median of 7 h after the last episode of airway obstruction or hypoxemia while breathing room air in an unstimulating environment gross-2006. Those are consultant survey medians rather than evidence, they belong to a superseded document, and the 2014 update dropped them. Quoting them as current ASA guidance is wrong in two ways at once.

The familiar STOP-Bang banding needs the same treatment. Only the top of the scale is measured: a score of 5 to 8 identifies a high probability of moderate-to-severe apnea chung-2012-stopbang. The paper that derived the questionnaire gives its cut-off for STOP alone, two or more affirmative answers, not for STOP-Bang chung-2008-stop. The 3 to 4 intermediate band that residents are taught traces to a narrative review, which this project does not admit as a source, so treat that band as convention rather than as evidence.

At the bedside

Identify the patient at increased perioperative risk from OSA before they reach the PACU, and treat the discharge decision for that patient as an obligation with an undefined end point rather than a timer gross-2014. The question to ask is not how many hours, but what would tell you the patient is no longer at risk of respiratory depression, and whether the setting you are sending them to could detect it if you are wrong.

The practical test the guidance points at is the room-air observation it does describe: watch the patient in an unstimulated environment, preferably while asleep, breathing room air, and see whether the saturation holds gross-2014. When you write the disposition, write which guideline you are applying and what its qualifier is. A resident who documents that consultants strongly agree with the plan is being accurate; a resident who documents that evidence supports the timing is not gross-2014.

Local practice differs. the monitored setting used after PACU discharge for patients at increased risk from OSA, the monitoring modality, and the criteria for ending it differ between institutions; follow local protocol and know which guideline it cites.

Check yourself

Three items. Every option carries an explanation. Progress is not saved.