Junger et al. Anesth Analg 2001; 92:1203-9

Reviewed by: R. Prasad, MD

Summary:

• Their computer system (AIMS) collects perioperative data (intraop, PACU, demographic, medical history) including vitals, drugs given
• Data on 27,626 pts; ~11K for evaluation, rest for validation.
• PONV prophylaxis: by MD preference. No 5-HT3 antagonists
• PONV treatment: metoclopramide, droperidol, dimenhydrinate
• Pts asked repeatedly if nauseated; medicated if true ... med record used as estimation of PONV incidence.
• Looked at long list of variables related to the patient (but no h/o PONV or motion sickness), the surgery, the anesthetic (meds used, ...), and PACU (duration, analgesics, clonidine)

• Incidence 7 - 8%

• Logistic regression equation: PONV probability in PACU = 1/(1+ e^ (4.9682 -0.8949*Female +0.6294*Smoker +0.917*PropofolMAC -0.8078*N2O -1.4301*IntraopOpioids +0.005*Age -0.0051*DurationMinutes)
• Accuracy 76% (ROC curve)
• Accuracy similar to Apfel and Sinclair
• N20 in combination with desflurane (but not propofol) increased PONV

• Yet another formula to play with on the Palm
• Unfortunately, did not include potentially important historical info (h/o PONV, motion sickness) because their database did not record this in any standardized way. Such a history may add 10-20% to the risk for PONV.
• Only looked at data in PACU - would like to follow patients out to 24 hrs.
• It would be nice if someone could find a statistically valid way of combining the data from some of these huge studies that have done logistic regression analysis on a large number of risk factors! Here's my attempt, which I developed using the very rigorous statistical method of copying and pasting: PONV Incidence (just for fun)