Process Analysis in Outpatient Knee Surgery: Effects of Regional and General Anesthesia on Anesthesia-controlled Time

The analysis of our ACLR clinical pathway database was approved by the Institutional Review Board for the Health Sciences, University of Pittsburgh Medical Center, Pittsburgh, Pennsylvania.

We queried our institutional database for consecutive patients undergoing ACLR by one surgeon over a 3-yr period (July 1995 through June 1998). Historical control data (June 1995 through June 1996) were compiled from existing medical records and hospital databases. Data from July 1996 through June 1998 were gathered within 2–4 weeks after the day of surgery and examined in the course of quality control and follow-up. The three anesthesia techniques compared were GA, GA–RA, and RA. Data were excluded from patients for whom postoperative hospital admission was preapproved or based on unplanned extension of the surgical procedure.

For time interval definitions, we used descriptions from the Association of Anesthesia Clinical Directors’ Procedural Times Glossary  6or terminology previously described in the literature. 1,7ACT consists of OR entry until surgical preparation begins plus the end of the surgical procedure until OR exit. Turnover time (TOT) is defined as the interval from “prior patient out of room to succeeding patient in room for sequentially scheduled cases.”6ACT and TOT are shown in figure 1.

Because not all surgical pavilions undergo uniform routing processes for their patients, this section outlines the steps patients went through on the day of surgery at our institution. After admission to the same-day surgery unit, patients were brought to the preoperative holding area when summoned by the OR control desk. The OR control desk was instructed to call for the next patient either by the OR nursing staff or by the anesthesia team. Throughout the 3-yr period of study, patients were summoned to the holding area sufficiently early to minimize delays between cases. On arrival to the holding area, patients were checked in by a team consisting of a recovery room nurse and a unit clerk. Holding area staff paged the responsible anesthesiologist, who interviewed the patient and obtained informed consent for the anesthesia care plan. The staff anesthesiologist was responsible for intravenous line insertion before OR entry, but may (rarely) have delegated this task to the “hands-on” anesthesia provider (e.g. , nurse anesthetist) after the latter was finished with the previous case in the OR. Intravenous insertion may also have been delegated to the RA resident, who was also available throughout the study interval to perform RA techniques under the medical direction of the attending anesthesiologist.

Once patients had intravenous access secured, the staff anesthesiologist was then able to provide preoperative sedation or initiate any preoperative RA technique with the rotating RA resident. The holding area nurse (one of the recovery room nurses) monitored patients after intravenous sedation or regional technique delivery; the nurse operated under the medical direction of an attending anesthesiologist who was immediately available. Likewise, the RA resident (who was not assigned to any specific OR) was immediately available for monitoring after regional procedures were performed. The view of all patients (and their monitors) was unobstructed to all holding area staff once intravenous sedation was given or RA techniques performed.

All patients waited in the holding area until their OR was ready. In the OR, GA and GA–RA patients had routine monitors applied and GA was induced, at which time the surgeons proceeded with preparation and positioning. RA patients also had monitors applied after OR entry, but with surgical anesthesia already activated, the surgeons were able to start preparation and positioning immediately.

The preoperative holding area times and estimated anesthesia procedure times (per anesthesia technique) are listed in table 1. Although preoperative time taken to perform RA techniques before OR entry is by definition not part of ACT, 1,6the extent to which OR entry was delayed by the preoperative performance of a regional technique is reflected in an increased TOT. In our group practice, any preoperative regional technique for the next patient was most commonly completed before the current patient was finished with surgery. With this in mind, the reader must understand that ACT does not necessarily capture anesthesiologist or anesthesia care team labor intensity with preoperative patients if preparations for the next patient are completed before the current patient is finished with surgery.

After surgery, patients were transported to either phase 1 recovery, consisting of the traditional postanesthesia care unit (PACU, with a 1:2 nurse-to-patient staffing ratio), or phase 2 recovery, which was the same-day surgery recovery unit (with a 1:6 nurse-to-patient staffing ratio). Because the PACU bypass fast-tracking criteria later described by White and Song 8were not available at the time of observational study, we used PACU bypass criteria as detailed in  appendix 1.

Our teaching hospital uses an anesthesia care team, in which nurse anesthetists or anesthesiology residents provide anesthesia services under the medical direction of the staff anesthesiologist. Anesthesiology residents rotating on the RA service are also medically directed by the staff anesthesiologist in the administration of preoperative techniques. The anesthesiology resident assigned to an OR does not perform preoperative RA techniques because of inherent time constraints in a busy outpatient surgery center.

The staffing model used in the ORs discussed in this study did not consist of staff anesthesiologists personally delivering anesthesia. Therefore, the anesthesia care team (staff anesthesiologist medically directing nurse anesthetists) was a staffing constant throughout the 3-yr period included in the database query.

 Appendix 2summarizes the specific anesthetic techniques used during the care of these patients. In our center’s clinical practice, patients were allowed to choose the anesthetic plan from the options available. Variability in anesthesia care options was minimized by an agreement among a limited number of staff anesthesiologists assigned to these ORs to follow an anesthesia care protocol consisting of the three categories mentioned (GA with or without femoral nerve block, or one of several “straight” regional techniques). Using this approach, patients are offered a choice of anesthesia techniques (when applicable), but more exact and less variable anesthesia process inputs result once the care technique is selected.

Symptom (pain, postoperative nausea, or vomiting [PONV]) incidence, associated nursing interventions (per patient), and discharge times observed with each anesthesia technique category were tracked. In addition, unplanned hospital admissions as a result of refractory pain or PONV were tracked and categorized for each anesthesia technique.

Differences in ACT were analyzed using two methods. The first was one-way analysis of variance, which was used to compare the three anesthetic techniques over the entire 3-yr period of study, irrespective of time. In this analysis, differences among anesthetic categories were determined post hoc  using the Bonferroni correction (P < 0.05 was considered significant). The second analysis of ACT was via  general linear modeling using a stepwise approach to determine which variables predicted ACT, specifically including chronological, seasonal, and autocorrelation effects. The variables included in this analysis included anesthetic technique, case sequence in particular day of study, case sequence throughout entire study period, day of week, month of year, and 3-month epoch of study. Autocorrelation analysis revealed a significant relation between the ACT of the current case and those of the five previous cases. The autocorrelation effect was incorporated in the modeling process by creating “lag variables” reflecting the ACT values of the first through fifth previous cases.

The objective of the modeling equation for ACT was to determine the extent to which anesthesia technique (GA, GA–RA, or RA) and other concurrent processes and variables (e.g.,  time of study) contributed to differences in ACT. In a linear modeling equation, all significant variables considered simultaneously would individually reflect P < 0.05. Variables with lower P  values in any model are more significant contributors to the equation studied.

Once all variables were identified for analysis, they were used to create a factorial regression model. This served as the starting point for a backward stepwise modeling technique, serially eliminating the least significant variables at each step. When the model was left with only significant terms, the minimal factorial regression models for ACT were considered complete.

One-way analysis of variance was used to compare the Association of Anesthesia Clinical Directors’ Procedural Times Glossary  definition of TOT values per year and per technique. 6Because no significant differences among techniques were seen, we then added TOT values to the ACT values per patient and performed general linear modeling. The objectives, analysis, and interpretation for the TOT plus ACT linear modeling equation was otherwise identical to the modeling described previously for ACT alone.

The incidence of postoperative pain and PONV, and the need for unplanned admission, are dichotomous (yes–no) variables. We defined pain and PONV as any symptom requiring a nursing intervention with a parenteral medication. For these comparisons based on anesthesia technique, the chi-square test was used. The number of nursing interventions needed to treat pain, PONV, and discharge times, are continuous variables. For these comparisons based on anesthesia technique, one-way analysis of variance was used, with the Bonferroni adjustment for multiple comparisons. For all tests, P < 0.05 was considered significant.

Data from 413 consecutive patients undergoing ACLR by one surgeon over a 3-yr period (from July 1995 through June 1998) were analyzed from our institutional database. During the 3-yr period, 44 of these patients were excluded from this analysis because of a preapproved hospital admission or admission because of unforeseen complications of the surgical site (e.g.,  bleeding, drainage). There were no significant demographic differences among the 44 admitted patients versus  the 369 patients ultimately pooled for analysis.

Table 1outlines the demographic data relevant to the 369 patients undergoing ACLR during the period of study from July 1995 through June 1998. This patient population was typically young and had few coexisting diseases, and their surgery was commonly necessitated by a sports-related injury.

Table 2lists the observational data associated with the three anesthesia technique categories. Time spent in the preoperative holding area was estimated based on 2 yr of recorded data by the holding area staff. Time spent in patient preparation by the anesthesiologist incorporates the attending and RA resident working together for regional techniques, but the attending working alone for GA preparation.

RA was associated with the lowest ACT (11.4 ± 1.3 min, mean ± 2 SEM). GA–RA (15.7 ± 1.0 min) was associated with lower ACT than GA used alone (20.3 ± 1.2 min). Annual (irrespective of technique) and per- technique (irrespective of time) differences in ACT and TOT are shown in table 3. Annual reductions in ACT were noted, as was an increase in TOT between the first and second years of the study. TOT values in the third year of study showed a significant decrease compared with TOT values of the previous 2 yr.

Per-technique differences in ACT showed that this time value was lowest with the use of a preoperative induction area (RA). Predictors of ACT differed with general linear modeling based on whether or not the TOT value was added to the ACT value. In both equations, the use of a preoperative induction area (RA) predicted the lowest ACT, whereas the use of in-OR preparation time (GA) predicted the highest ACT.

Two predictors of ACT (other than anesthetic technique) were time- series related: patient sequence in the database relative to the 3-yr chronology of the study, and 3-month epoch in which the patient had surgery. One other predictor was seasonal (month of the year). Generally speaking, ACT decreased throughout the 3-yr period and was lowest during the months of January, March, April, and December. There was no autocorrelation effect present in the final model of ACT. The modeling results are shown in table 4.

One predictor of ACT plus TOT was seasonal (month of the year); another was time-series related (3-month epoch in which the patient had surgery). As would be expected, ACT plus TOT also decreased over time and was lowest during the months of January, March, August, and December. An autocorrelation effect was present in the final model of ACT plus TOT: the ACT plus TOT values for the lag-2 and lag-3 cases (i.e. , the second and third cases, respectively, before the index case) were noted to be significant predictors. These modeling results are shown in table 5.

Operating room managers and surgeons typically express concern about TOT as a surrogate for ACT. However, examining TOT as an isolated indicator of process efficiency does not sufficiently address ACT. TOT is influenced by many contributors to the perioperative process, including OR nursing, surgery, anesthesia, housekeeping, and preoperative and postoperative nursing. As a result, TOT may not reflect process improvements of one discipline if other departments’ processes are not simultaneously improved. The importance of distinguishing TOT from ACT is well-illustrated in our data: in 1996–1997, the shift in anesthetic technique from in-OR preparation GA (predominantly) to preoperative preparation (frequently epidural techniques) was associated with a decreased ACT but an increase in TOT. In 1997–1998, our response to the increased TOT was to summon patients to the preoperative holding area sooner. This led to TOT values that were lower in 1997–1998 than at any time previously. The sum of TOT and ACT was lower in 1997–1998 than it was in the previous 2 yr. Although we did not quantify the processes of all disciplines, it appears that TOT was influenced by improvements in the turnover process and not just by anesthesia-related processes.

The room turnover study by Mazzei 7analyzed what we are now defining as ACT plus TOT. In perioperative environments where effective room turnover process management transcends departmental barriers and all disciplines function effectively as a team, we feel that the anesthesia team’s contributions to the time when surgeons are not operating is best captured by ACT plus TOT. Our ACT plus TOT data reflected autocorrelation, which is likely the result of the influences of anesthesia processes and those of other disciplines on room turnover. In perioperative environments where prolonged turnovers can be attributed to the inefficiency of one or several disciplines (but not necessarily the anesthesia service), anesthesia contributions to surgical case delays are best captured by ACT (without TOT added). We noted no autocorrelation in the prediction of ACT, which may indicate that preoperative interventional anesthesia processes using appropriately designed staffing models need not contribute to prolongation of TOT (or ACT) in the surgical care process.

Routine, short-duration, relatively noninvasive procedures that are most prudently managed with only one type of anesthetic (e.g. , GA for pediatric myringotomy) will likely have a narrow range of ACT values. However, we found that invasive outpatient orthopedic procedures benefit, albeit marginally, from a preoperative induction area. Additionally, RA techniques provide faster emergence and postoperative transport to the PACU. Additional labor intensity for GA at these milestones prolonged TOT, ACT, or both; if emergence and PACU transport are prolonged, then ACT plus TOT values are increased.

For example, if a patient required a PACU stay after GA (as opposed to bypassing PACU after RA), the sign-over to PACU delayed the anesthesia provider’s return to prepare the next patient for OR entry. This sequence prolonged TOT (and thus prolonged ACT plus TOT). When the next patient also underwent GA induction in the OR with no preoperative preparation in an induction room (i.e. , no preoperative RA technique), then the time until surgeons begin preparation and positioning is prolonged. Finally, at the conclusion of the GA, the time to extubate and exit the OR is prolonged compared with RA or GA–RA.

The timing of RA techniques in the induction room had varying effects on ACT and TOT. All anesthetic preparation performed intraoperatively before surgical preparation and positioning or after the end of the surgical procedure (such as with GA) prolonged ACT. If the start of a preoperative block in an induction room is delayed until after the previous patient is finished, then both ACT and TOT are prolonged.

Regional techniques performed preoperatively and producing surgical anesthesia (e.g.,  epidural or lumbar plexus and sciatic nerve blocks) prolonged ACT plus TOT if the finish time for the RA in the induction room did not coincide with or precede the time when the OR was ready, and if an operative patient required the physical presence of the anesthesiologist after the surgical procedure was finished (e.g.,  extubation after GA) but before the conclusion of the RA on the following preoperative patient.

ACT plus TOT was minimized when anesthesia preparation in the induction room for the subsequent patient was completed by the OR ready milestone (e.g. , with epidural anesthesia or lumbar plexus and sciatic nerve block anesthesia). Timely completion of anesthesia preparation for the subsequent patient was more likely when the current patient in the OR did not require the anesthesiologist’s presence for emergence (i.e. , GA was not used for the current patient in the OR).

This study examined one anesthesia staffing model—that of the anesthesia care team and a RA resident, operating under the medical direction of an attending anesthesiologist, working in a OR suite operating at greater than 80% capacity. Although we did not address staffing costs in this study, we can speculate that in a large-volume teaching hospital with available staff that can be designated for regional block placement (e.g.,  RA resident), and medically directed hands-on providers (e.g.,  nurse anesthetists), this staffing model would produce desirable time reductions in the OR. In addition, the associated outcomes with using RA (e.g.,  symptom reductions, avoided unplanned admissions) may ultimately justify such a staffing expenditure. However, this study was not rigorously designed (in a prospective, randomized, or blinded fashion) to study patient symptoms and outcomes. Until properly designed prospective randomized studies are performed and published, the associated outcomes based on anesthesia technique remain uncertain.

The cost profiles of physician-delivered anesthesia versus  the care team model would differ significantly, as would the time available during or between cases to perform preoperative RA techniques. Indeed, staffing alterations, or computer simulations of staffing models, would serve as an interesting basis of subsequent study.

This study examines one surgical procedure by one surgeon operating in two ORs. If the staffing model described (emphasizing RA) were applied to an eight-OR suite doing fairly uniform procedures operating at full capacity with each staff surgeon operating in only one OR, much greater time savings could be realized. Given the following assumptions: the first-case preparation and position begins at 7:30 a.m. and the desired finishing time is 3:00 p.m., there is no variability in surgical case time (or ACT), and ACT plus TOT values are 30, 35, and 40 min for RA, GA–RA, and GA, respectively, we would expect the following results. If each of the eight ORs had four 90-min cases (from beginning preparation and positioning to procedure finished), the RA rooms would finish at 3:00 p.m., the combined GA–RA rooms would finish at 3:15 p.m., and the GA rooms would finish at 3:30 p.m. If the OR suite were operating at full capacity, the time saved would still be insufficient to schedule an additional surgical procedure, but personnel costs for forced overtime over the course of 1 yr may be reduced. Quantification of such savings and comparison of other staffing models necessitate further study (e.g.,  using computer simulation modeling).

The most common postoperative complaints after outpatient orthopedic surgery are PONV and pain. Although very common causes of unplanned admissions, PONV and pain are potentially preventable by the anesthesiologist. Anesthesia care plans that are designed in isolation to speed patients through the surgical process and minimize ACT or TOT without considering important symptoms and outcomes is not rational. The anesthetic techniques used in this study were designed not only to decrease ACT and TOT but also to minimize complications that lead to unplanned admissions.

Gold et al.  9demonstrated that PONV was the most common anesthesia-related complication forcing unplanned admission after same-day surgery. Carroll et al.  10demonstrated that PONV was a costly in-hospital complication in surgical outpatients. This group also demonstrated that patients could experience PONV on homegoing up to 35% of the time even when not admitted, creating notable patient distress. 11The in-hospital incidence of PONV after GA in our study (with or without femoral nerve block) was 36%, which was consistent with the postoperative incidence noted by Carroll et al.  10,11This high PONV incidence after GA occurred despite using antiemetic prophylaxis for most patients from July 1996 through June 1998. Our shift of care emphasis from GA to RA has been associated with significant reductions in PONV before discharge (14%;P < 0.001 by the chi-square test). Further study of this patient population is necessitated to determine the incidence of PONV on homegoing. Additionally, more extensive multimodal approaches for the prevention of PONV after GA were not specifically addressed.

Postoperative pain is another common symptom, especially after invasive reconstructive knee surgery. Table 2demonstrated the decreases in the incidence of pain and PONV and the required nursing interventions for these symptoms associated with the use of RA. We believe that observational data analysis such as that in table 2is important because improving ACT and TOT while concomitantly decreasing nursing labor requirements is the only method that will ensure full capture of theoretical cost savings. With the OR operating at full capacity, ACT reductions may decrease the costs of forced overtime of all perioperative personnel, whereas reducing nursing labor requirements in the postopera-tive period (through preemptive symptom management) may reduce the costs of postoperative nurse staffing. More rigorous study designs or statistical simulations are necessitated to confirm this theory.

Future studies of this patient population could reasonably hypothesize that the use of GA in the absence of multimodal preemptive therapy would be associated with an increased risk of PONV. Future studies of femoral nerve block analgesia could hypothesize that its use would be associated with lower incidence of postoperative pain and thus fewer unplanned admissions. Additionally, we did not specifically examine the role of preoperative interventions for GA (with induction rooms). Although this approach is widespread throughout much of Europe, it has not found wide acceptance in the United States. Therefore, we chose to compare preoperative interventions with RA, a more standard practice in the United States. Further studies of this type must address the role of preoperative intervention, or preparation with GA and RA, to successfully draw conclusions regarding the role of ACT and TOT between these two modalities.

For OR management purposes, minimizing ACT and TOT is certainly in the best interest of the surgical care team. 12However, designing anesthetic interventions only to minimize either factor may adversely affect short- and intermediate-term quality indicators of patient outcome, such as unplanned admissions. In our patient population, the exclusive use of intraoperative RA, perioperative administration of femoral nerve block analgesia, or both was associated with fewer unplanned admissions because of PONV and pain; these findings necessitate further prospective study.

Consecutive patients receiving preoperative interventions (RA, femoral nerve block analgesia, or both) were associated with the lowest ACT, the lowest sum of ACT plus TOT, and the lowest incidence of unplanned hospital admission. Both of these processes and outcomes are desirable features of ambulatory anesthesia. This study cannot answer whether the extra anesthesia staffing costs in the described staffing model (vs.  OR suites operating below capacity or with more variable case mixes) would be offset by sufficient cost savings (by reducing nursing labor intensity or unplanned admissions).

TABLE 

TABLE