Anesthesia groups may wish to decrease the supervision ratio for nontrainee providers. Because hospitals offer many first-case starts and focus on starting these cases on time, the number of anesthesiologists needed is sensitive to this ratio. The number of operating rooms that an anesthesiologist can supervise concurrently is determined by the probability of multiple simultaneous critical portions of cases (i.e., requiring presence) and the availability of cross-coverage. A simulation study showed peak occurrence of critical portions during first cases, and frequent supervision lapses. These predictions were tested using real data from an anesthesia information management system.
The timing and duration of critical portions of cases were determined from 1 yr of data at a tertiary care hospital. The percentages of days with at least one supervision lapse occurring at supervision ratios between 1:1 and 1:3 were determined.
Even at a supervision ratio of 1:2, lapses occurred on 35% of days (lower 95% confidence limit = 30%). The peak incidence occurred before 8:00 AM, P < 0.0001 for the hypothesis that most (i.e., >50%) lapses occurred before this time. The average time from operating room entry until ready for prepping and draping (i.e., anesthesia release time) during first case starts was 22.2 min (95% confidence interval 21.8-22.8 min).
Decreasing the supervision ratio from 1:2 to 1:3 has a large effect on supervision lapses during first-case starts. To mitigate such lapses, either staggered starts or additional anesthesiologists working at the start of the day would be required.
What We Already Know about This Topic
The most appropriate ratio of anesthesiologists to providers would avoid lapses of supervision during critical portions of anesthetic cases. A simulation study suggested this occurs most commonly with simultaneous first starts.
What This Article Tells Us That Is New
In a review of 1 yr of data from a tertiary hospital, lapses occurred commonly during first-case starts even with a 1:2 supervision ratio.
These data suggest that either staggered starts or additional anesthesiologists working at the start of the day would be needed to reduce lapses during critical periods.
ANESTHESIOLOGISTS often function in anesthesia care teams (e.g. , supervising concurrently two or more certified registered nurse anesthetists).1,–,7Many anesthesia groups perceive an incentive to decrease their supervision ratio.8,–,10Because a ratio lower than 1:2 does not satisfy accreditation requirements of the American College of Graduate Medical Education, ratios lower than 1:2 apply to nurse anesthetists, not anesthesia residents.‡Because many hospitals focus on tardiness of first-case starts11,12and offer many such starts,13,–,16anesthesiologist staffing is sensitive to the supervision ratio.
The number of operating rooms (ORs) that an anesthesiologist can supervise is limited by the probability of occurrence of two or more simultaneous events (i.e. , critical portions) requiring either physical presence or a time-sensitive, nonpreemptive interaction. The probability of supervision lapses is also influenced by the availability of other anesthesiologists to cross-cover. The consequence might be limited to a case delay, but patient safety could be affected when there are coincident critical physiologic events.
In the United States, invoicing Medicare for professional anesthesia services requires that the anesthesiologist “personally participates in the most demanding procedures in the anesthesia plan, including induction and emergence, where indicated.”§However, to reduce the risk of substandard care,17many institutions do not reveal patient insurance information. Consequently, all patients are supervised in accordance with Medicare rules. Furthermore, anesthesiologists' time before induction likely will increase from implementation of the World Health Organization surgical safety checklist.18
Paoletti and Marty19used simulation to estimate the risk of a supervision lapse in surgical suites with various numbers of ORs (2–18) performing a mix of elective cases of various durations (0.8–4.5 h) and a range of anesthesiologist supervision ratios (1:1, 1:2, 1:3). Their model parameters were based on data from several French hospitals. The simulated risk of a supervision lapse peaked at the start of the day. Risks ranged from 14% to 87% for inability to supervise all critical portions of cases at a 1:2 ratio, depending on case length (higher with shorter cases) and the size of the suite (lower with more ORs). Increasing the supervision ratio to 1:3 markedly increased the risk. Providing an unassigned “floater” anesthesiologist greatly reduced the risk.
We explored predictions of the French simulation study using real data captured from an anesthesia information management system to determine the incidence and timing of simultaneous critical portions of cases.
Our first hypothesis was that, as predicted,19on one-third of days, there would be supervision lapses even with a supervision ratio of 1:2.
Our second hypothesis was that, as predicted,19the peak incidence of supervision lapses occurred at the start of the day (e.g. , not during lunch breaks). If true, a supervision ratio less than 1:2 would require an increase in first-case start delays; first-case starts staggered sufficiently to allow the later first case to start on schedule20; additional anesthesiologists available at the start of the day; or anesthesiologists not present for all critical portions of cases.
If the first and second hypotheses were true, then the mean anesthesia release time would determine the average delay when two patients, supervised by the same anesthesiologist, were simultaneously ready for induction and all other anesthesiologists were occupied. We previously published how to use such mean times for anesthesia group economic analyses of first-case starts.12,13
Our third hypothesis was that anesthesia release times for first-case starts would average 22 min, in the midrange of values determined at Yale-New Haven Hospital.21
Materials and Methods
After Thomas Jefferson University Institutional Review Board (Philadelphia, Pennsylvania) approval with waiver of informed consent, we reviewed all 15,656 records in the hospital's anesthesia information management system on nonholiday weekdays between May 3, 2010 and May 1, 2011‖that took place in the 24 ORs comprising the two largest surgical suites. Inpatient and outpatient procedures are performed in these suites, but not cardiac surgery or diagnostic gastrointestinal procedures. The times of events and descriptive information listed in table 1were obtained. Heart rate, oxygen saturation, and invasive and noninvasive blood pressure values were retrieved from the anesthesia information management system database, recorded at 1-min intervals. Actual room locations where procedures took place were determined as previously described.22
We considered the anesthesia providers (i.e. , those individuals delivering direct anesthesia care) to be busy during the interval from the beginning to the end of anesthesia. The duration of breaks and lunch relief was considered as the interval from the documented start of the break to the documented end of the break, or lasting the mean duration of documented breaks if only the start time of the break was recorded in the anesthesia information management system, which is typical practice (72% of cases) for our providers. Where the end time of the break was not documented, the mean lunch break duration (30 min, based on 1,998 documented breaks) was substituted (presumed for breaks occurring between 11:00 AM and 1:30 PM, which is when lunch is offered). For breaks outside this period with a missing end time, the duration was set at the mean duration of such breaks (i.e. , 15 min, based on 2,776 documented breaks).
Each day was divided into 1,440 1-min intervals, during each of which the total number of providers who were busy was determined. We considered anesthesiologists to be occupied in tasks that cannot be preempted (i.e. , unable to leave the patient being cared for) during the periods listed in table 2. For each day, the number of anesthesiologists who were occupied as specified was determined during each 1-min interval.
Table 3lists the physiologic events (hypoxemia, hypotension, and hypertension) considered critical portions of cases. The physiologic event definitions were based on published manuscripts demonstrating adverse outcomes and represent prolonged alarm conditions, as opposed to transient or false alarms. The duration of each such event corresponded to when the threshold for the critical event occurred (e.g. , after 10 min with systolic blood pressure less than 70 mmHg), until when the alarm trigger no longer was in effect (e.g. , systolic blood pressure ≥70 mmHg). The events we included deliberately underestimated the critical portions of cases to take a conservative approach with respect to the incidence of supervision lapses, increasing the chance of rejecting Hypothesis 1 (discussed in the Statistical Methods section). For example, a blood pressure of 220/140 lasting 20 min during a case scheduled for 1 h was not classified as a critical physiologic event in our analysis, although such instances would almost certainly trigger a call to the supervising anesthesiologist. The same goes for a systolic blood pressure of 75 in a patient undergoing carotid endarterectomy, or a progressive drop in oxygen saturation measured by pulse oximetry from 100% to 90% in a patient undergoing robotic prostatectomy. Our approach was also conservative because there are other physiologic perturbations where the anesthesiologist would likely be notified that we did not include (e.g. , ST segment depression, hypercapnia not responding to an increase in minute ventilation, or runs of supraventricular tachycardia). In addition, we did not include “false alarm” conditions (e.g. , disconnection of an electrocardiogram electrode, kinking of the blood pressure tubing, or plugging of the capnograph sampling tubing) that may generate a call to the attending to help troubleshoot and/or resolve the problem.
For each minute of the day, we determined the total number of critical portions of cases that occurred simultaneously (fig. 1). For example, if at 8:40 AM there was a patient being extubated, a patient ready for induction of general anesthesia, and a patient with hypoxemia due to severe bronchospasm, there would be three critical portions of cases in the interval from 8:40:00 AM to 8:40:59 AM. Consequently, the total number of providers needed would equal the number of ORs with cases running plus three anesthesiologists.
For each minute of each workday excluding Thursdays, the running minimum number of anesthesia providers during overlapping 5 min was calculated (i.e. , to determine the number of ORs with cases). Thursdays were excluded because the OR starts 1 h later on this day and we were assessing supervision as a function of time of day. Over the same overlapping intervals, the minimum number of simultaneous critical portions of cases was calculated (i.e. , to determine the number of anesthesiologists needed). For each workday, the number of ORs was calculated as the maximum of the running minimums of the number of simultaneous providers. The number of anesthesiologists needed daily was the maximum of the running minimums of simultaneous critical portions of cases. The ratio of the number of ORs to number of anesthesiologists needed was then calculated for each day. This was most commonly simply 24 ORs divided by the maximum number of anesthesiologists needed for at least 5 min. For hypothetical ratios from 1.0 to 3.0 (i.e. , one anesthesiologist supervising from one to three ORs), the percentage of workdays for which the daily ratio was smaller was calculated. The use of overlapping 5-min intervals deliberately resulted in underestimation of this ratio (i.e. , increasing the chance of rejecting Hypothesis 1). For the ratio of 2.0, the lower 95% confidence limit was calculated for the percentage of workdays for which at least one supervision lapse would have occurred. The 95% confidence interval (CI) was calculated using the method of Blyth-Still-Casella (StatXact-9, Cytel Software Corporation, Cambridge, MA).
For each minute of each of the 202 workdays, excluding Thursdays, the total number of providers needed was calculated = provider in the operating room + anesthesiologist (if a critical portion of a case occurred) + and person on break (if applicable). Next, for each workday, the minute of the day with the largest total number of providers was calculated. That minute was then classified as “first case” if it occurred at 8:00 AM or earlier, otherwise “morning” if before 10:56 AM, otherwise “lunch” if before 1:31 PM, and otherwise “afternoon.” We calculated the percentage of days for which a minute at or before 8:00 AM had the largest total number of providers for the day, along with the 95% lower confidence limit. We tested whether the percentage exceeded half (i.e. , most) of the days. The calculations were performed twice, once with ties for the time of the day being assigned to the earlier time of day and once to the later time of day. For example, if the daily maximum of 35 anesthesia providers were needed on a day both at 7:58 AM and at 8:02 AM, then first the maximum would be attributed to the 7:58 AM “first case” and next attributed to the 8:02 AM “morning.” The calculations were also repeated using anesthesiologists' critical portions instead of the total number of providers needed.
For all combinations of the 253 workdays and OR first cases of the day, the time from each OR entrance to anesthesia release was known from the anesthesia information management system data. The probability distribution of the n = 5,769 times to release were not normally distributed with or without inverse squared, inverse, inverse square root, logarithmic, square root, or squared transformations of the release time durations (all Lilliefors tests P < 0.00001, Systat 13, SYSTAT Software, Chicago, IL). Therefore, the mean was taken for each day. The 253 means followed a normal distribution (Lilliefors test P = 0.42). The means had neither statistically significant Pearson auto-correlation from 1 day to the next (Pearson r = −0.01, P = 0.94) nor from 1 week to the next (r = 0.11 P = 0.08). Therefore, the 95% two-sided CI for the mean release time was calculated using the Student t distribution, with the sample size being the 253 workdays. Similarly, the overall mean was compared with the anesthesia release time of 22 min determined at Yale-New Haven Hospital21using Student one group two-sided t test.
Hypothesis 1: Staffing Lapses
The percentage of days during which there would have been at least one 5-min interval with too few anesthesiologists to supervise all critical portions of cases at varying ratios of ORs to anesthesiologists is shown in figure 2. Even at a ratio of 1:2, there would have been at least one such lapse in supervision for 35% of days (lower 95% confidence limit = 30%). At a ratio of 1:3, there would be supervision lapses on 99% of days (lower 95% confidence limit = 96%).
Extrapolating from figure 5b of the French simulation study19with 24 ORs, a staffing ratio of 1:2, and one additional floater anesthesiologist (i.e. , effective supervision ratio of 1:1.8), the expected incidence of supervision lapses is 12%. We observed a 12% incidence with a supervision ratio of 1:1.7.
The first hypothesis that supervision lapses would take place on one-third of days and that our results would be similar to the simulation study was confirmed.
Hypothesis 2: Time of Day with Largest Number of Providers Needed
The average peak activity (total providers needed) during cases occurred at the start of the workday for most days (fig. 3, table 4, P < 0.0001). This was especially true for critical portions of cases (i.e. , times that would influence anesthesiologist staffing; table 3). The second hypothesis was confirmed.
Hypothesis 3: Anesthesia Release Time
The mean number of minutes of critical portions of first-case starts was 22.2 min (95% CI 21.8–22.8 min, SD 2.8 min). This observation matched observational findings reported previously from Yale-New Haven Hospital21(P = 0.29). Thus, the third hypothesis that the mean number of critical minutes for first-case starts would match the anesthesia release time measured by observers21was confirmed.
Effect of Providing Higher Supervision Ratios or Staggered First-case Starts on Supervision Lapses
Because the three hypotheses were satisfied, as a sensitivity analysis, we examined the effect on supervision lapses of either lowering the supervision ratio from 1:2 at the start of the day to 1:3 after first cases had begun or supervising at a 1:3 ratio throughout the day with staggered first-case start times. The former strategy would be possible only if there were anesthesiologists with nonclinical assignments (e.g. , academic institutions), whereas the latter approach could be instituted anywhere. When critical portions of cases occurring at or before 8:00 AM and breaks were excluded, at least one supervision lapse would occur on 14% of days at the 1:3 supervision ratio (95% lower confidence limit = 10%). However, when breaks were included, supervision lapses increased to 62% of days (95% lower confidence limit = 56%; fig. 4). The breaks affecting the maximum supervision ratio were principally lunch reliefs (see fig. 2and table 4). These findings indicate that at a 1:3 supervision ratio, additional providers (e.g. , certified registered nurse anesthetists) would be needed to provide breaks. In contrast, if supervision were maintained at 1:2 throughout the day, there would be supervision lapses on only 0% and 2% of days, excluding and including breaks, respectively. Thus, additional providers would not be necessary at a 1:2 supervision ratio. Overall, the financial benefit of decreasing the supervision ratio from 1:2 to 1:3 is offset by the need for additional nonanesthesiologist providers.
In this study, we confirmed results of the French simulation study,19showing that even at a supervision ratio of one anesthesiologist for every two anesthesia providers, all simultaneous critical portions of cases could not be supervised on one-third of days without occasionally waiting for the anesthesiologist. We also confirmed that the largest number of providers is needed at the start of the day, and that is also when there was the highest incidence of critical portions of cases. The mean anesthesia release time (22 min) we measured was close to that measured at Yale-New Haven Hospital.21That time represents the average expected delay in starting the second case when an anesthesiologist has two patients who are ready for induction simultaneously and there is not another anesthesiologist who is available to cross-cover.
Our findings and the simulation results19are in contrast to the study of Wright et al. ,23which found that cases with a start time after 3 PM had the highest proportion of adverse events. We obtained different results because our focus was on the time of the day with the largest total number of critical portions among all ORs. Wright et al. 23considered when each individual case had the highest risk.
Administrators who want to reduce their anesthesia group's costs24by encouraging them to decrease their anesthesiologist supervision ratios need to consider the effect of our findings on the timeliness of first-case starts, which is often a major institutional focus.11,12At a ratio of one anesthesiologist to three anesthesia providers, it will not be possible to start all ORs simultaneously and have sufficient anesthesiologists to supervise all critical portions of cases on most days. Either the administrators will need to accept the fact that the additional OR often will be delayed from its scheduled start time, or agree to rearrange the OR schedule so that first cases supervised simultaneously by each anesthesiologist will have staggered start times.20The former approach can lead to discontent, because such delays are publicly visible.25The use of staggered starts has a built-in expectation that some ORs will start later than other ORs. For some organizations this may be advantageous (e.g. , surgeons running multiple ORs or who simply prefer to start somewhat later than the “official” start time may embrace this change). Provided the ORs selected for the staggered start times20are those with the most expected underutilized OR time, this has no economic disadvantage.12,13,26,27
Another potential approach to the problem of supervision lapses during first cases of the day is for the anesthesia group to make additional anesthesiologists available at the start of the day. Then, once the ORs have been started, some of these individuals are released to perform other duties important to the department (e.g. , research, informatics, and management and administrative duties). The importance of Hypothesis 2 is in knowing that lunch breaks are not the bottleneck; rather, it is the first case starts that must be considered economically.12,24However, the importance of our sensitivity analysis is in showing that this approach then necessitates adding additional nonanesthesiologists for breaks, which may nullify the economic benefit.
The fact that some organizations do not routinely provide breaks is not a limitation of our study to such practices, because our results of the importance of the start of the workday with respect to the peak incidence of staffing lapses would then be even stronger . Similarly, the fact that we studied a tertiary hospital with many long cases rather than an outpatient surgery center with short cases is not a limitation because, from the simulation study,19our results would be even stronger for short cases. Instead, the principal limitations of our study relate to the definitions of critical portions of anesthetics. Although we relied on process times recorded in an anesthesia information management system, such times recorded by nurses in an operating room information system could be used equivalently, as shown by Sandberg et al. 28
During our analysis, we assumed, as did Paoletti and Marty,19that any anesthesiologist can go into any OR when a critical portion of the case occurs and provide supervision equivalent to the anesthesiologist who is otherwise occupied and cannot be interrupted. If complex patients are involved or an extended discussion about management has taken place, such substitution may provide suboptimal patient care. To the extent that all anesthesiologists are not equivalent and thus not able to supervise every critical portion of cases (e.g. , a patient to receive a regional block that the available anesthesiologist does not feel qualified to perform), the percentage of days with a lapse in supervision with a 1:2 supervision ratio would be even larger than the observed 35%.
There are aspects of our analysis related to our definitions of critical portions of cases (tables 1and 2) that could result in some readers viewing our conclusions as too conservative. Several of our colleagues offered feedback that they do not think that it is necessary for the supervising anesthesiologist to be physically present for induction or emergence in straightforward cases with experienced certified registered nurse anesthetists, as long as they are immediately available. The extent to which anesthesiologist presence is required during and soon after the anesthesia release time varies highly among countries because of varying regulatory requirements and within countries among institutions (e.g. , depending on local requirements for participation in the preoperative briefing). Because the intraoperative briefing including the surgeon and all anesthesia providers reduces mortality,18likely its inclusion will be increasingly prevalent.
In summary, we showed that the start of the OR day is the period of time when the anesthesiologist supervision requirement is greatest. Even with lunch breaks included, this result is so robust that changes in the anesthesiologist supervision ratio can be described to administrators simply in terms of the effect on first-case starts. This finding is useful because the psychology of first-case starts is already understood (e.g. , how they are interpreted economically).11Decreasing the supervision ratio by anesthesiologists from 1:2 to 1:3 will have a great effect on the timeliness of the start of the first cases of the day due to the high incidence of simultaneous critical portions of cases peaking at that time. As the economics of first-case starts are also fully developed, the decision to stagger first- case starts appropriately11,–,13,26,27,versus having more anesthesiologists can be modeled for each facility.11,12,24Unless one of these options is chosen, the consequence will be a marked increase in the incidence of supervision lapses.