The final logistic regression model contained 5 input variables, which included patient, surgical, and anesthetic factors. Patient factors included BMI ≥ 40 kg/m2 and the presence of hypertension. Surgical factors included the surgical subspecialty and scheduled case duration. The anesthetic factor included the utilization of general anesthesia. The final multivariable logistic regression model with each variables’ β-coefficient, OR (95% CI), and P value are listed in Table 4. The strongest predictors for extended PACU length of stay were scheduled case duration (for every increase of 1 hour, OR 1.60, 95% CI 1.41–1.83, P < .0001) and utilization of general anesthesia (OR 2.20, 95% 1.51–3.20, P < .0001) as the primary anesthetic when compared with MAC. The variables that are most protective for extended PACU length of stay included surgical/procedural specialties such as ophthalmology (OR 0.11, 95% CI 0.06–0.022, P < .0001) when compared with general surgery.
The model has an AUC of 0.754 (95% CI 0.733–0.774) on the training set and 0.722 (95% CI 0.698–0.747) on the test set, with no difference between the 2 ROC curves (P = .062) (Figure 1). Calibration, as indicated by goodness-of-fit by the HL test, is shown in Figure 2. The logistic regression model had good calibration for the data in both the training and test data set as indicated by a P value of >.10 (P = .211 and P = .719, respectively). Supplemental Digital Content 1 and 2 (Supplemental Appendix Tables 1 and 2, http://links.lww.com/AA/B607, http://links.lww.com/AA/B608) lists the results of the HL test for the training and test data, respectively.
To demonstrate utilization of the model, a simple example is provided here. The example patient is an 85-year-old woman with morbid obesity and hypertension, who is scheduled to undergo a laparoscopic cholecystectomy (General Surgery), which is booked for 3 hours. She will undergo general anesthesia as her primary anesthesia type. The median length of stay in the PACU at her facility is 45 minutes with a third interquartile range of 90 minutes. Her probability of having a PACU length of stay greater than or equal to 90 minutes from the logistic regression model will be calculated as:
Probability of extended PACU stay = 1/(1 + e−(−2.426 + 0.515 + 0.377 + 0.789 + 0.472 × 3)) = 0.662
We developed a predictive model for extended PACU length of stay for outpatient surgeries using multivariable logistic regression that demonstrated adequate discrimination and calibration. The predictors in the model included presence of hypertension, morbid obesity, primary anesthesia type (general anesthesia), scheduled case duration, as well as surgical specialty. The goal of this model was not to serve as a tool to decrease PACU length of stay, but rather to identify surgical patients at higher risk for this outcome metric before day of surgery during the operational decision-making period.
Several studies have identified clinical and administrative predictors for prolonged PACU length of stay in various surgical populations.2–10 When studying outcomes such as PACU length of stay, it is useful to focus predictive modeling on special subsets, rather than all surgical cases (ie, all adult patients undergoing general anesthesia). By focusing on a special group, in this case outpatient procedures, bias introduced from the wide range of surgical subspecialties may be reduced. Furthermore, past studies included variables that are not known preoperatively, such as events that either occurred intraoperatively or during postoperative care (ie, intubated state in the recovery room, intraoperative arrhythmias, and postoperative pulmonary symptoms).2,3,6,9 The covariates we chose to be included in our model included only those that are known preoperatively. Of note, we utilized scheduled case duration (time allotted for surgery before actual case) as opposed to the actual case duration. This is because actual case duration is not known until the postoperative period and therefore would not be useful in this type of predictive modeling.15
There is no uniform definition of what is considered “extended PACU length of stay” in current studies; therefore, results may not be transferrable across studies or institutions. For example, the following definitions have been used: duration of stay > 2 hours16 or duration greater than 60 minutes.2,17 National data, consisting of hundreds of institutions, were used in another study; however, the definition of prolonged PACU length of stay was heterogeneous across health care facilities.10 Finally, other studies treated PACU length of stay as a continuous variable and the predicted durations were compared.4,7,9 In the current predictive model, extended PACU length of stay was treated as a binary outcome defined as time greater than third interquartile range. Because of the heterogeneity of PACU workflows across institutions, we felt that defining the outcome this way would be generalizable. In this sense, one is comparing the outcome metric within an institution’s own constraints.
Patient-specific, anesthetic, and surgical covariates were found to be statistically significant in its association with this outcome metric. It is not surprising that among primary anesthesia type, general anesthesia, when compared with MAC, was associated with extended PACU length of stay. This is likely related to the nature and complexity of surgeries requiring general anesthesia than those that do not. Furthermore, general anesthesia requires much more anesthetic medication and usually longer periods of recovery postoperatively than MAC. Longer scheduled case duration was associated with extended PACU length of stay. Case duration has been noted in previous studies to be associated with this outcome.2,3,10 The assumption here is that longer anesthesia could be related to more complex surgeries and increased anesthetic utilization, thereby increasing recovery time. The presence of obstructive sleep apnea was significantly associated with the outcome in univariable analysis but was no longer significant in the multivariable model. Morbid obesity, however, remained significant, which is a subpopulation with increased risk for sleep apnea. These patients are at increased risk of postoperative airway obstruction, especially in the setting of opioid utilization.18 Upper airway problems, especially from airway obstruction due to pharyngeal laxity from pharyngeal muscle weakness, are one of the more common PACU complications.19 It is clinically prudent to observe these patients relatively longer than patients without obstructive sleep apnea. The presence of hypertension revealed increased odds for extended PACU length of stay likely related to the hemodynamic challenges faced in this population during recovery.
Many surgical subspecialties, including Otolaryngology, Gastroenterology, Ophthalmology, and Plastic Surgery had decreased odds for extended PACU length of stay when compared with General Surgery. This could be related to the nature of surgeries and the patient populations in specialties such as General Surgery and Gynecology, as well as the fact that these subspecialties afford higher risks to PONV in outpatient surgery. For example, Sarin et al20 demonstrated that, in the ambulatory surgery suite, laparoscopic cholecystectomy, tubal ligation, and pelviscopy had increased odds for PONV.
Previous studies have found that age was associated with extended PACU length of stay, which is not consistent with the current findings.4,10 Although the geriatric population may be more prone to a number of postoperative complications and may require longer periods of time for adequate recovery compared with their younger counterparts, age alone, as suggested by this study, does not necessarily predict extended PACU length of stay. Likely, it could be the associated comorbidities related to age that contribute to it. Patients appropriate for outpatient surgery usually have less comorbidity burden and undergo less complex surgeries; therefore, this sample population likely consists of healthier elderly patients.
Operating room management is a complex process that involves multiple planning steps at different time points: (1) strategic decision making, which requires years of planning before implementation and involves permanent changes to operating room functionality; (2) tactical decision making, which is performed once or twice a year and tends to change operating room capacity, for example, through changes in total block time; (3) operational decision making, which has a time course of weeks to minutes and involves allocation of operating room time, scheduling of cases, and staff assignment; and finally (4) decisions made the day of surgery based on maintaining patient safety, operating room efficiency, and patient waiting times.21 Predicting cases that may afford prolonged recovery room length of stays may be useful during the operational decision-making stage, in which scheduling of cases and staff assignment occurs (specifically, in this case, recovery room staffing).
Minimizing the peak number of patients in the PACU is important because this can reduce the number of PACU admission delays (which occurs when there is inadequate nurse staffing in the PACU to safely take care of the incoming patients). Furthermore, prolonged PACU length of stay at the end of the day may potentially overutilize nursing if needed staffing is not adequately estimated. If given historical information regarding inflow into the PACU from the operating rooms and a set timeframe at which nursing shifts are made, the staffing of the recovery room can be adjusted.22,23
However, what is the optimal sequence at which to schedule cases? The answer requires many interrelated variables and its usefulness is dependent on the type of operating room environment.23–25 The concept of resequencing cases does not afford much benefit, in terms of optimizing staffing assignment, in large PACUs24; however, it does have benefit in facilities with just a few operating rooms, such as that of outpatient surgery centers (as is performed in the current study).23,25 In facilities with a small number of operating rooms (such as ambulatory surgical suites), the goal here is to minimize the number of recovery room nurses needed in a given day by finding the optimal sequence in which cases should be scheduled that would minimize the peak number of patients in the PACU and minimize makespan (ie, the completion time of the last patient in the PACU).23,25 Although the current study does not attempt to create such an algorithm, which is beyond the scope of this article, it does create a model that helps identify patients that may stay in the recovery room longer. This potentially could be an important variable inputted into future algorithms that can calculate optimal case sequencing in outpatient surgery centers with a small number of operating rooms in an effort to minimize makespan. For example, would it be advantageous to resequence cases such that patients with predicted prolonged PACU length of stay are scheduled earlier in the day? Of note, this predictive model does not forecast a specific time that the last patient of the day exits the PACU because this will also depend on the actual duration of cases – information only known by the end of the surgical day. Still, it is unclear whether identifying at-risk patients in advance will minimize makespan, not until case- sequencing algorithms that put this variable into account are tested.
There are several limitations to this study. One is its retrospective design. In addition, there are some nonclinical factors that may contribute to extended PACU length of stay and were not included in this model. These include transportation issues, nursing availability for education/discharge, and physician availability for orders.26,27 These types of delays are difficult to predict and are the very issues that may be avoided with appropriate predictive modeling and subsequent surgical and resource planning/utilization. It is unclear how these administrative or logistic delays would bias the results, because there has been no study performed at our hospital identifying whether certain patient or surgical characteristics are associated with this particular type of delay. Certainly, these types of delays may be a reason why the discriminating ability is not closer to 100%. Another limitation is that there are a large number of possible comorbidities that may be included in model building – many of which are not included here. We chose the most common comorbidities for this population. Plus, certain comorbidities have varying degrees of severity (ie, degree of systolic function in congestive heart failure) that were not controlled for here. However, at our institution, patients are selected to be appropriate for outpatient surgery based on severity of disease, so this mitigates this problem. Finally, this is a single-institution experience. However, the patient population and surgical mix may be similar to those of other outpatient surgery practices. Our next steps are to validate this model externally.
In conclusion, we developed a predictive model with excellent discrimination and goodness-of-fit that can help identify patients at higher odds for extended PACU length of stay. These predictions may aid in the development of effective case-sequencing methodologies aimed at optimizing recovery room staff scheduling in outpatient facilities with a relatively small number of operating rooms.
Name: Rodney A. Gabriel, MD.
Contribution: This author helped design the study, conduct the study, collect the data, analyze the data, and prepare the manuscript.
Name: Ruth S. Waterman, MD.
Contribution: This author helped design the study, conduct the study, analyze the data, and prepare the manuscript.
Name: Jihoon Kim, MS.
Contribution: This author helped analyze the data and prepare the manuscript.
Name: Lucila Ohno-Machado, MD, PhD.
Contribution: This author helped design the study, conduct the study, analyze the data, and prepare the manuscript.
This manuscript was handled by: Richard C. Prielipp, MD.
1. Weissman C. The enhanced postoperative care system. J Clin Anesth. 2005;17:314–322.
2. Kim WO, Kil HK, Kang JW, Park HR. Prediction on lengths of stay in the postanesthesia care unit following general anesthesia: preliminary study of the neural network and logistic regression modelling. J Korean Med Sci. 2000;15:25–30.
3. Waddle JP, Evers AS, Piccirillo JF. Postanesthesia care unit length of stay: quantifying and assessing dependent factors. Anesth Analg. 1998;87:628–633.
4. Seago JA, Weitz S, Walczak S. Factors influencing stay in the postanesthesia care unit: a prospective analysis. J Clin Anesth. 1998;10:579–587.
5. Corey JM, Bulka CM, Ehrenfeld JM. Is regional anesthesia associated with reduced PACU length of stay?: A retrospective analysis from a tertiary medical center. Clin Orthop Relat Res. 2014;472:1427–1433.
6. Edler AA, Mariano ER, Golianu B, Kuan C, Pentcheva K. An analysis of factors influencing postanesthesia recovery after pediatric ambulatory tonsillectomy and adenoidectomy. Anesth Analg. 2007;104:784–789.
7. Silins V, Julien F, Brasher C, Nivoche Y, Mantz J, Dahmani S. Predictive factors of PACU stay after herniorraphy in infant: a classification and regression tree analysis. Paediatr Anaesth. 2012;22:230–238.
8. Bothner U, Georgieff M, Schwilk B. The impact of minor perioperative anesthesia-related incidents, events, and complications on postanesthesia care unit utilization. Anesth Analg. 1999;89:506–513.
9. Chung F, Mezei G. Factors contributing to a prolonged stay after ambulatory surgery. Anesth Analg. 1999;89:1352–1359.
10. Gabriel RA, Wu A, Huang CC, Dutton RP, Urman RD. National incidences and predictors of inefficiencies in perioperative care. J Clin Anesth. 2016;31:238–246.
11. Chung F. Recovery pattern and home-readiness after ambulatory surgery. Anesth Analg. 1995;80:896–902.
12. Dexter F, Tinker JH. Analysis of strategies to decrease postanesthesia care unit costs. Anesthesiology 1995;82:94–101.
13. Lasko TA, Bhagwat JG, Zou KH, Ohno-Machado L. The use of receiver operating characteristic curves in biomedical informatics. J Biomed Inform. 2005;38:404–415.
14. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988;44:837–845.
15. Dexter F, Dexter EU, Ledolter J. Statistical grand rounds: Importance of appropriately modeling procedure and duration in logistic regression studies of perioperative morbidity and mortality. Anesth Analg. 2011;113:1197–1201.
16. Lau LL, Hung CT, Chan CK, et al. Anaesthetic clinical indicators in public hospitals providing anaesthetic care in Hong Kong: prospective study. Hong Kong Med J. 2001;7:251–260.
17. Brown I, Jellish WS, Kleinman B, et al. Use of postanesthesia discharge criteria to reduce discharge delays for inpatients in the postanesthesia care unit. J Clin Anesth. 2008;20:175–179.
18. Mulier JP. Perioperative opioids aggravate obstructive breathing in sleep apnea syndrome: mechanisms and alternative anesthesia strategies. Curr Opin Anaesthesiol. 2016;29:129–133.
19. Hines R, Barash PG, Watrous G, O’Connor T. Complications occurring in the postanesthesia care unit: a survey. Anesth Analg. 1992;74:503–509.
20. Sarin P, Urman RD, Ohno-Machado L. An improved model for predicting postoperative nausea and vomiting in ambulatory surgery patients using physician-modifiable risk factors. J Am Med Inform Assoc. 2012;19:995–1002.
21. Dexter F, Epstein RH, Traub RD, Xiao Y. Making management decisions on the day of surgery based on operating room efficiency and patient waiting times. Anesthesiology 2004;101:1444–1453.
22. Dexter F, Epstein RH, Penning DH. Statistical analysis of postanesthesia care unit staffing at a surgical suite with frequent delays in admission from the operating room–a case study. Anesth Analg. 2001;92:947–949.
23. Dexter F, Epstein RH, Marcon E, de Matta R. Strategies to reduce delays in admission into a postanesthesia care unit from operating rooms. J Perianesth Nurs. 2005;20:92–102.
24. Marcon E, Dexter F. An observational study of surgeons’ sequencing of cases and its impact on postanesthesia care unit and holding area staffing requirements at hospitals. Anesth Analg. 2007;105:119–126.
25. Hsu VN, de Matta R, Iee CY. Scheduling patients in an ambulatory surgical center. Nav Res Logist. 2003;50:218–238.
26. Cowie B, Corcoran P. Postanesthesia care unit discharge delay for nonclinical reasons. J Perianesth Nurs. 2012;27:393–398.
27. Tessler MJ, Mitmaker L, Wahba RM, Covert CR. Patient flow in the post anesthesia care unit: an observational study. Can J Anesth. 1999;46:348–351.
Supplemental Digital Content
© 2017 International Anesthesia Research Society