The total cost to a hospital of an individual drug can be high, either because the drug itself is expensive, or because it is used for many patients. Although the cost of anesthetic drugs is only a small proportion of the total cost of most procedures,1 inhaled anesthetics such as sevoflurane appear commonly in the top 10 drugs by costs for our hospital pharmacy.
Most volatile anesthetics are delivered from an out-of-circuit plenum vaporizer. The actual consumption of these drugs is then determined by the rate of fresh gas flow (FGF) through the vaporizer and the vaporizer dial setting. In many delivery systems, it is difficult to determine actual vapor consumption, and surrogate measures, such as the weight of the vaporizer before and after a case,2–4 the reported flow rate in the middle of a case,5 or estimated use across a mix of anesthesiologists and cases6 may be used. Some newer anesthetic delivery systems, such as the Datex ADU, incorporate electronic flow transducers and electronic control of vapor delivery. With these systems, it is possible to record actual patterns of FGF, allowing new insights into inhaled anesthetic consumption. These systems were first introduced into our hospital in 1999 at about the same time that limitations on the minimum FGF that can be used when administering sevoflurane were removed by the New Zealand Government drug regulation bodies.
We conducted audits of anesthetic FGF rates during 20017 and 2006. The audits during 2001 were part of a study to assess the impact of education on use of FGF rates. The decision to audit in 2006 was precipitated by several factors. The cost of anesthetic drugs forms a major part of our hospital pharmacy budget and we sought evidence to support continuing access to the more expensive drugs. Additionally, we had introduced a locally developed inhaled anesthetic delivery guidance system.8 This system uses FGF rates, vaporizer settings, and end-tidal vapor concentrations derived from the anesthesia machine and patient monitor to provide forward predictions of end-tidal and effect-site vapor concentrations.9 The near real time operation (10-s maximum lag) of the system allows the user to speculatively modify vapor delivery and see the consequence of that action on end-tidal and effect-site vapor concentration. The “Predictor,” as it is colloquially known, takes the form of a computer screen mounted on top of the anesthesia machine with a compact personal computer mounted in the lower part of the machine’s chassis. This system has been shown to increase the speed and precision of control of sevoflurane when making step changes in end-tidal sevoflurane to achieve defined targets.8
All data were collected in accordance with the guidelines of the New Zealand Health Information Privacy Code. The 2001 data were collected following the guidelines of the Canterbury Regional Ethics Committee; the 2006 data collection was approved by the New Zealand Multi-Region Ethics Committee.
In 2001, we conducted audits of FGF, which are described in detail elsewhere. The initial audit period of March 2001 was not advertised and the data collection system was sufficiently discrete as not to alert the anesthesiologist. After the first audit period, the department was instructed in the theory, practice, and economic benefit of FGF reduction. In addition, written material was made available and individual tutoring to trainees in the techniques was offered. The second audit period was performed during November 2001 and a prominent notice was placed on the anesthetic machine advising that a FGF audit was in progress.
The collection of audit data used a locally developed system where data are collected from the anesthetic monitor (Datex A/S3) and the anesthetic machine (Datex ADU) into a computer placed on the base of the machine. The computer has a slimline “pizza box” style case (Macintosh LC or Quadra). Data are requested every 10 s from the ADU and monitor using a locally developed application. From the ADU data stream, we are able to extract flow rates of oxygen and nitrous oxide and air; vapor in use and vaporizer settings; and total FGF. The system automatically restarts after any interruption of power supply. During the audits in March and November 2001, no keyboard or monitor was attached to the computer. During the 2006 audit, the predictive system described above was in place in the operating rooms (OR) being audited. The predictor system possesses the data logging attributes of the original system, having been developed in part from it. However, this retained feature was not known other than to the authors. No notice of the audits conduct was given.
In 2001 data were collected for two 1-mo periods in a single OR (Theater A). This room is used for elective general and vascular surgery during the normal working week and as the primary location for acute general and vascular surgery out of hours.
The 2006 data were collected from four ORs over a 3-mo period; Theater A, (no change in the pattern of use); Theater B—-elective, in hours, general surgery with occasional out of hours use; Theater C—-acute general and vascular surgery during normal working hours; Theater D—-acute orthopedic. Most patients anesthetized in these rooms are adults. All anesthesia is provided by anesthesiologists or anesthesia residents. As our interest is in the overall behavior of the population of anesthesia providers rather than individuals, and because of the logistic complexity, no attempt was made to ascribe particular cases to individual practitioners.
Data were transferred from the computer in each location and run through an analysis program which counted each of the 10-s samples. For this audit, only those time periods when the vaporizer setting was above zero were considered. Data were sorted into 12 bins by total FGF for each volatile anesthetic and for total FGF. Cumulative totals of gas use with isoflurane, sevoflurane, or desflurane were calculated to allow calculation of the mean gas flow with each drug. Samples were combined by summing the counts in corresponding gas flow and vapor bins and calculating means and frequency distributions from this combined data set.
Frequency distribution curves were compared using a two sample Kolmogorov-Smirnov Test.* This test was chosen because it compares the entire cumulative distribution curves. Comparisons were made between the 2001 samples, between the pooled 2001 results and the pooled 2006 results, and between the pooled 2006 result and that for each individual OR. A P value <0.01 was taken as significant.
Data collection took place over 66 days in 2001 and 72 days in 2006. Volatile anesthetics were in use for 439 h in 2001 and 505 h 2006.
The mean FGFs for the three time periods (Mar 2001, Nov 2001 and 2006) were 1.95, 2.10, and 1.27 L/min respectively. The median value occurred in the 1.0-1.5 L/min bin for both 2001 periods and in the 0.5-1.0 L/min bin in 2006.
Figure 1 shows cumulative distribution curves comparing the three sample periods. The frequency distribution for the pooled 2001 data was significantly different from that for the pooled 2006 data (D = 0.39, P < 0.01). There was no statistically significant difference in the frequency distribution curves between the 2001 samples (D = 0.07, ns) or between the mean of the 2006 samples and each individual room (OR A, D = 0.04; OR B, D = 0.16; OR C, D = 0.06; OR D, D = 0.14) as shown in Figure 2.
Between 2001 and 2006, there was an increase in the use of sevoflurane and desflurane, and a corresponding large decrease in isoflurane use (Table 1).
This study has revealed a 35% decrease in FGF rates from a mean of 2.0 L/min in 2001 to 1.3 L/min in 2006. Both sets of FGF rates are in a range many would consider acceptable, representing a balance between the economy of reducing flow rates and the complexity (real or imagined) of true low flow or, at the limit, closed circuit anesthesia.
This study is useful and unique because the data were collected automatically and in real time. Most previous studies of anesthetic consumption rely either on user reports,5,6 bulk long-term measures, such as total hospital consumption of durgs,10 or cumbersome techniques, such as weighing of vaporizers before and after a case.2–4 Although this last method is considered accurate, it is not practical for routine use.
Major cost saving occurs by reducing “high” flows to flows in the 1-2 L/min range.5,10 The magnitude of the change seen in this study (0.7 L/min) appears small; however, this 35% decrease represents a considerable cost saving given that more than 90% of our volatile anesthetic use is with either sevoflurane or desflurane. The total cost of sevoflurane at Christchurch Hospital for the 2005-2006 financial year was $US360,000. Simulations based on models we have used previously8 suggest that to maintain an effect-site level of 1 MAC sevoflurane between 20 and 30 min after the start of an anesthetic requires 55 mL/min of sevoflurane vapor at a FGF of 2.0 L/min. This reduces to 40 mL/min at a flow of 1.3 L/min. Using these values as an approximation for the effect of flow rate changes on vapor consumption over an entire case, if gas flows had been maintained at the level seen in 2001, then sevoflurane usage and therefore costs would have been 55/40 of that actually seen in 2006. This translates to an estimated total cost of $495,000, which is more than $130,000 over the actual cost. Thus, although the change in flow rate is small, this degree of change spread across a large number of cases can add up to a significant saving in consumption and therefore cost.
Although a major focus of most health systems is cost minimization, there are other, possibly more important, reasons to reduce FGF rates and total volatile anesthetic consumption. These include the potential for reduced workplace and environmental pollution, both from production of these anesthetics and from release into the atmosphere.
There are some limitations of this study. It has focused only on changes in FGF use with volatile anesthetics but does not give us any information about other possible changes in anesthetic practice, such as the use of IV or regional techniques. There are significant differences among the study periods, in particular spreading the observations across a number of ORs. This has potentially altered the case mix, the acuity of cases and the mix of staff between specialist and trainee. However, taking this wider approach gave us a broader picture of practice than that provided by a single location, and the similarity of results suggests this is a reasonable summary of current practice in our hospital.
It is interesting to speculate on the reasons for the changes observed over the past 5 years. These changes may represent a culture shift (or perhaps culture drift) whereby flow reduction below 2 L/min has become the norm rather than the domain of the enthusiast. This is supported by the large proportion of time with flow rates <1 L/min.
One explanation for these changes could be changes in staffing. We have an extremely stable specialist population with 90% of the senior staff employed at the time of the original audit still present during the second audit. The resident staff has completely changed over this time period and the number of specialist anesthesiologists has grown by 30%, recruited in large part from our own trainees. We have no specific data to address individual attitudinal factors among staff, but one can consider the staff population as a whole. It would seem unlikely that within a large, relatively stable department there has been a shift in “personality” sufficient to account for the change in practice seen. It seems most likely that we are seeing a change in behavior rather than a change in personnel. Although information about low flow anesthesia might account for this change, our original audit does not support this.7
The tools which are likely to facilitate flow reduction are now commonplace, and possibly better understood. These include agent analyzers, which have been mandated for more than 10 yr and the availability of modern delivery systems, such as the Datex ADU, which provide delivery of reduced flows simple with accurate delivery systems, minimal leakage, and ventilation systems that compensate for gas flow changes. In 2001, these were still relatively new, and available in 6 of 11 ORs. In 2006, they were in all 17 ORs.
In addition, we had introduced our volatile effect-site/predictive display system into routine practice during the period between the two audits. The evolution of the system into the current form, and also its gradual introduction to different anesthetizing locations, has meant that a formal pre- and post-introduction audit process would be problematic. The system had been in a stable “final release” form in the audited locations for more than a year before the 2006 audit. As well as its predictive functions, the system displays the current vapor consumption both as mL/min of vapor and as $/h.
We cannot attribute the reduction in FGF to the presence of our predictive system. However, the manner in which the system presents data may represent a form of the individualized feedback that has been suggested as necessary to maintain changes in FGF.5 Certainly, the system provides continuous feedback to the user in a manner that is immediately understandable. Our system is unable to provide data for individual providers, but does give us a good overview of practice. We believe that the utility of collecting anesthesia delivery information in this way should encourage developers of anesthesia record keeping systems to include FGF and vapor delivery information in both display and audit functions.
This study demonstrates that clinical practice with regard to FGF rates has not been static in our department, but has shown an improving economy of use. This sort of study is important when demonstrating the responsible use of anesthetics in the setting of ever increasing costs. The cause for the decrease in FGF rates is not addressed in our study but may reflect the availability of improved technology. It would be valuable if this type of study could be replicated in an environment before and after an equipment upgrade.
1. Demeere JL, Merckx C, Demeere N. Cost minimisation and cost effectiveness in anaesthesia for total hip replacement surgery, in Belgium? A study comparing three general anaesthesia techniques. Acta Anaesthesiol Belg 2006;57:145–51
2. Struys MM, Kalmar AF, De Baerdemaeker LE, Mortier EP, Rolly G, Manigel J, Buschke W. Time course of inhaled anaesthetic drug delivery using a new multifunctional closed-circuit anaesthesia ventilator. In vitro comparison with a classical anaesthesia machine. Br J Anaesth 2005;94:306–17
3. Enlund M, Wiklund L, Lambert H. A new device to reduce the consumption of a halogenated anaesthetic agent. Anaesthesia 2001;56:429–32
4. Coetzee JF, Stewart LJ. Fresh gas flow is not the only determinant of volatile agent consumption: a multi-centre study of low-flow anaesthesia. Br J Anaesth 2002;88:46–55
5. Body SC, Fanikos J, DePeiro D, Philip JH, Segal BS. Individualized feedback of volatile agent use reduces fresh gas flow rate, but fails to favorably affect agent choice. Anesthesiol 1999;90:1171–5
6. Tohmo H, Antila H. Increase in the use of rebreathing gas flow systems and in the utilization of low fresh gas flows in Finnish anaesthetic practice from 1995 to 2002. Acta Anaesthesiol Scand 2005;49:328–30
7. Kennedy RR, French RA. An audit of anaesthetic fresh gas flow rates and volatile anaesthetic use in a teaching hospital. NZ Med J 2003;116:U438
8. Kennedy RR, French RA, Giles S. The effect of a model based predictive display on the control of end-tidal sevoflurane during low flow anesthesia. Anesth Analg 2004;99:1159–63
9. Kennedy RR. The effect of using different values for the effect site equilibrium half time on the prediction of effect site sevoflurane concentration—-a simulation study. Anesth Analg 2005;101:1023–8
10. McKenzie A. Reinforcing a “low flow” anaesthesia policy with feedback can produce a sustained reduction in isoflurane consumption. Anaesth Intens Care 1998;26:371–6
*http://home.ubalt.edu/ntsbarsh/Business-stat/otherapplets/ks.htm.© 2008 International Anesthesia Research Society