Morbidity after anesthesia is common, but clinical trials have demonstrated that various adverse events associated with anesthesia can be reduced. However, in anesthetic practice, there have been few studies which have attempted to actively alter anesthetic management and measure changes in patient outcome .
Postoperative pain which leads to severe patient discomfort is a common problem in anesthesia practice, and can be associated with increased respiratory complications and length of hospital stay [2,3]. Therefore, anesthetic strategies which reduce postoperative pain, if routinely used, could enhance overall patient care.
The primary objective of this study was to change anesthesiologists' practice of pain management by introducing directed interventions based on the principles of academic detailing combined with personalized feedback profiles [4,5]. Secondary objectives were to determine whether there was any effect on the rate of early postoperative pain (either in the postanesthesia care unit [PACU] or within the first six-hour interval after PACU discharge) or on other indicators of morbidity (postoperative respiratory problems) and resource use (length of PACU and hospital stay).
Ethics approval was obtained for this study of directed interventions and perioperative data collection at two hospitals. A before-after study design with a concurrent control group was chosen. Baseline data were collected over a 6-mo period and compared to three subsequent 6-mo periods (education, feedback 1, and feedback 2). At the study hospital, educational and personalized feedback interventions, reports displaying each anesthesiologist's rate of use of the promoted strategies and their patients' level of postoperative pain, were introduced after the baseline data collection. At the control hospital data were collected but no active interventions were given. To determine whether there was any long-lasting effect 1 yr after the interventions were stopped, at the study hospital data were collected for an additional 6-mo period (follow-up).
Both hospitals were tertiary-care teaching institutions affiliated with the University of Toronto, each fully staffed by board-certified anesthesiologists. The number of operating theaters and the yearly elective case load were similar at the two hospitals and neither hospital had an active pain management team at the time of the study. The age groups of anesthesiologists were distributed as follows: 16% were less than 40 yr and 13% older than 60 yr at the study hospital compared to the control hospital where 45% were less than 40 yr, and 10% older than 60 yr. An equal number of anesthesiologists were involved in critical care, 25% at the study hospital and 20% at the control hospital. At the initiation of the study all anesthesiologists at the study hospital were advised of the timetable of the study, the objectives of the research, and the intent to provide individual reports of practice and outcome. At the control hospital, anesthesiologists were aware of only the data collection and not the study goals.
For a 2-yr period, April 1992 to March 1994, all consecutive inpatients who had major gynecological, joint replacement or repair, limb amputation, and invasive renal and abdominal procedures requiring laparotomy were observed. Patients who did not have a general anesthetic and those transferred directly to an intensive care unit bypassing the PACU were not included in the study population.
We used methodologies which have been previously shown to be successful in changing physicians' practice patterns. These included academic detailing, a methodology to improve clinical decision-making with approaches used by pharmaceutical companies, and personalized feedback [4,5]. Prior to the start of data collection, three anesthesiologists at the study hospital, two PACU nurses, and the three investigators chose the patient strategies to be promoted based on current literature reviews, cost, availability, and ease of implementation. The preventive strategies chosen for promotion were patient-controlled analgesia (PCA), nonsteroidal antiinflammatory drugs (NSAIDs), epidural morphine (EPIMORPH), and regional nerve blocks. Morphine was the opioid of first choice for intravenous PCA and no background infusions were used. NSAIDs included those specifically ordered by the anesthesiologist as premedication and either indomethacin (rectal suppository) or ketorolac (intravenous or intramuscular) administered intraoperatively or postoperatively for the prevention of pain. As the study period was limited to 6 h after PACU discharge, only a single injection of epidural morphine was administered. Regional nerve blocks included wound infiltration, extremity nerve block, and epidural without opioid.
An "opinion leader," an anesthesiologist recognized as an initiator of new therapeutic modalities and as an influential leader, spearheaded the arrangements, timetables, and distribution of all literature at the study hospital. The director of pain management was designated the "academic influential." Educational interventions included six seminars organized to encourage active participation, written summaries, and distribution of published literature dealing with general aspects of pain [6-9] or specific treatment regimens [10-19]. The rate of postoperative pain among the study hospital's patients was also presented showing that there was considerable scope for improvement. Speakers included anesthesiology staff with expertise in pain management and external experts who had coordinated research studies or had initiated pain management services.
After a 6-mo educational period, individual personalized feedback forms were distributed to all anesthesiologists at the study hospital. Four personalized feedback forms were given at 12, 15, 18, and 22 mo after the commencement of the study. Each anesthesiologist's feedback report detailed the rate of use of the promoted patient strategies (specifically PCA, NSAIDs, EPIMORPH, and regional nerve blocks) for a specified length of time. The form also included the rate of excessive pain in the PACU (defined as moaning or writhing in pain or nursing care dominated by pain control) and pain scores in the early postoperative interval (during the first 6 h after PACU discharge as delineated on a visual analog scale) for each anesthesiologist's patients. The feedback forms also included comparisons to previous time periods and to the three anesthesiologists whose patients had the lowest rate of postoperative pain. All educational information and feedback forms were presented in a positive manner with emphasis on patient improvement. Anesthesiologists were free to choose one or more of the promoted strategies and were not guided by mandatory policies or patient protocols.
Perioperative Data Collection
Patient anesthetic records completed by the anesthesiologist included demographic and comorbid conditions, anesthesia techniques, airway management, and all drugs and doses used during the operation. Drug therapy in the PACU and PACU complications were identified by daily review of the PACU patient record which contained a list of potential patient complications including definitions (Appendix). This PACU record was used at both hospitals for 6 mo prior to the initiation of the study and PACU nurses had undergone training on the recognition and documentation of complications. All specific strategies (PCA, NSAIDs, EPIMORPH, and regional nerve blocks) initiated by anesthesiologists (either preoperatively, intraoperatively, or as preventive strategy in PACU) were documented on the anesthetic or PACU record.
Patients' experience of pain during the first 6 h after PACU discharge was determined from patient interviews and chart reviews completed the day after surgery. Research nurses who rotated between the two hospitals had been trained to interview and review charts in a standardized manner using mock sessions with the investigators. Patients were specifically questioned about pain in the incisional area with rest and during activity (movement or deep breaths). Patients scored the amount of pain for the entire 6-h interval using an 11-point scale with 0 indicating no pain, and 10 the worst pain ever . Questions about activity and any other relevant problems were also asked. After the interview, a chart review was completed to determine all drugs and doses used in this early postoperative interval and any additional patient problems (Appendix).
The study comprised five 6-mo time periods at the study hospital (baseline, education, feedback 1, feedback 2, and follow-up completed 1 yr later) and four (no follow-up) at the control hospital. For the primary objective, the anesthesiologist was the "unit of analysis." Only anesthesiologists who were present for the entire study period at each hospital were counted (24 anesthesiologists at the study hospital and 21 at the control hospital). The percentages of each anesthesiologist's patients who received at least one promoted strategy or a specific treatment were determined and compared over time across the two hospitals and for each hospital using a repeated-measures analysis of variance (P < 0.01). This method tests to see whether there are any significant statistical "interactions" between independent variables (i.e., hospital and a specific time period) which will indicate whether there has been a differential change in the proportion of patients receiving a pain management strategy over time across the two hospitals. To determine the magnitude of the change in practice, an unpaired t-test was used to compare the percentage of each anesthesiologist's patients (mean +/- SE) who received either any preventive strategy or a specific preventive strategy for each time period compared to the baseline period (P < 0.01).
Next, we used the patient as the "unit of analysis." The rate of use of the four promoted strategies among all patients included in the study was calculated across the different time periods at both hospitals and compared using the chi squared statistic (P < 0.01). In addition, we used a z-test to compare the absolute changes in rates between the two hospitals .
For the secondary objective, we compared the rate of excessive pain in the PACU during the baseline period with all other study periods for each hospital (chi squared statistic). For the early postoperative interval after PACU discharge, patients who were not interviewed or able to answer questions were excluded from the analysis (<5% at both hospitals). Pain scores at rest and with activity after PACU discharge were displayed as means with 99% confidence intervals. The values for the different study periods were compared using the Kruskal-Wallis test (P < 0.01) .
To determine whether the changes in practice pattern which occurred over time were associated with any increase or decrease in morbidity or resource use, the rate of other patient PACU events and other patient problems after PACU discharge were compared to the baseline period using the chi squared statistic, Fisher's exact test, and unpaired t-test where appropriate (P < 0.01). Similarly, PACU length of stay (hours) and length of stay in hospital (days) were compared to the baseline period.
For each 6-mo time period there were at least 525 patients at the study hospital (total n = 3413) and at least 375 patients at the control hospital (total n = 1753). At the study hospital 67% of patients were female, 68% were more than 40 yr old, 86% were ASA I or II, 40% had gynecological procedures, 32% orthopedic procedures, and 22% open abdominal procedures. Patient case mix was similar at the control hospital. Characteristics of patients at the two hospitals over time were unchanged except for less major gynecological surgery at both hospitals, more abdominal procedures at the study hospital, and more arthroplasty procedures at the control hospital.
There was a significant change in the percentage of each anesthesiologist's patients who received any promoted strategy, both over time and between hospitals (repeated-measures, analysis of variance [ANOVA] hospital by period interaction P < 0.001) (Figure 1). At the study hospital over the 2-yr study period, there was a significant increase in anesthesiologists' use of any preventive strategy (repeated-measures ANOVA, period interaction P < 0.0001). At the control hospital, there was a smaller but significant increase in anesthesiologists' use of any preventive measure (repeated-measures ANOVA, period interaction P < 0.003).
At the study hospital increases were found in the percentage of each anesthesiologists' patients (mean value) who received PCA during all time periods compared with baseline and a further increase compared with the education period (P < 0.01) (Figure 2). In contrast, at the control hospital anesthesiologists' use of PCA was increased only in the feedback-2 period. Anesthesiologists' use of NSAIDs also increased at the study hospital, specifically compared to the baseline period (Figure 3); there was no increase at the control hospital. The use of any strategy as well as PCA and NSAIDs by anesthesiologists was maintained at the study hospital for the follow-up period. Changes in anesthesiologists' use of EPIMORPH and regional techniques were not significant at either hospital (Figure 4 and Figure 5).
With the patient as the "unit of analysis," comparing the feedback-2 period with baseline, the proportion of patients who received any promoted strategy increased from 0.7% to 45.6% (P < 0.01) at the study hospital and from 8.5% to 31.3% at the control hospital (P < 0.01) (Table 1). This absolute increase at the study hospital was greater than at the control hospital (44.9% vs 22.8%, P < 0.0001). During feedback 2, for patient PCA use, the absolute increase was 34.6% at the study hospital and 19.6% at the control hospital (P < 0.0001). NSAID use was increased by 14.4% at the study hospital but there was no change at the control hospital (P < 0.0001). Patients' use of EPIMORPH and regional techniques was infrequent, but there was a small but significant increase in the proportion of patients who received regional techniques at the control but not at the study hospital.
The rate of excessive pain noted in the PACU was unchanged between baseline and all subsequent periods at both hospitals (Table 2). At both the study and control hospitals, the mean pain scores at rest were unchanged for the feedback-2 period compared to the baseline period. However, mean pain scores with activity at both hospitals were decreased during the feedback-2 period (P < 0.01). This decrease was small at both hospitals.
Changes in indicators of morbidity and resource use between baseline and feedback-2 period are displayed in Table 3. For the study hospital rates of respiratory problems in the PACU and use of naloxone after PACU discharge were unchanged over time. The proportion of patients at the study hospital receiving postoperative morphine (intravenous or intramuscular) increased as did the total cumulative dose of morphine (from 0.27 +/- 0.14 to 0.37 +/- 0.27 mg/kg, P < 0.01). The cumulative dose of postoperative morphine over time was unchanged at the control hospitals. At the study hospital, both length of stay in PACU and in hospital were unchanged from baseline.
Pain management practices were successfully changed at the study hospital; after 18 months of education and individualized feedback, anesthesiologists' use of PCA and NSAIDs was increased compared to baseline. This improvement was sustained for one year after directed interventions had been discontinued. The absolute increase over time in the proportion of patients who received either PCA or NSAIDs was also greater at the study hospital. As well as an increased use of the promoted strategies, there appeared to be an additional effect on the awareness of pain as an important problem for patients. There was a significant increase in the use and dose (mg/kg) of morphine at the study hospital with no increased untoward respiratory events.
Despite the improvement in the anesthesiologists' use of pain management strategies, there were only modest improvements in patient outcomes; a small decline in mean pain scores with activity during the first six-hour interval after PACU discharge was found at the study hospital. The control hospital showed similar declines in postoperative pain scores so that any change in patient outcomes cannot be attributable to the directed interventions. Why were we not more successful at decreasing the rate of excessive pain?
Unfortunately, not all of the promoted strategies were readily available to anesthesiologists. Neither were they equally efficacious in all patients, and not all patients were treated. The use of PCA was initially restricted to certain nursing floors and only to patients who could be properly instructed. Other studies have noted that PCA may not actually reduce pain scores even though patient satisfaction was improved . The stock of NSAID suppositories (indomethacin 100 mg) in the operating room was unreliable, and some anesthesiologists noted that they were not comfortable with using NSAIDs in suppository form. Ketorolac (Toradol[R], Hoffmann-La Roche, Mississauga, Canada) was available, but several anesthesiologists expressed concern over its cost. In addition, there was apprehension by both anesthesiologists and surgeons related to the potential for increased bleeding and renal complications associated with the use of NSAIDs. Although NSAIDs in some studies have been associated with a reduction in opioid use and improvement in pain scores in the early postoperative period, not all studies are in agreement .
At both the study and the control hospitals, use of EPIMORPH was limited to patients discharged to step-down or critical care areas with higher nursing: patient ratios and monitors for pulse oximetry. Regional nerve blocks using long-lasting local anesthetics are effective, but lack of familiarity with technique may have influenced the anesthesiologist's decision in avoiding this strategy. Although other studies have shown superior pain relief using regional techniques with or without opioids, these modalities were used too infrequently in this study to improve the overall outcome of patients. Not all patients at the study hospital received the promoted strategies. In the feedback-2 period alone, only 35% of patients received PCA and only 14% received NSAIDs. In fact, 54% received none of the promoted strategies for pain management leaving considerable room for improvement.
Why did anesthesiologists' use of one of the promoted strategies, PCA, increase at the control hospital when these anesthesiologists received no active interventions? Although specific journal articles on pain management were highlighted, discussed, and distributed at the study hospital, anesthesiologists at the control hospital had access to the same journals. Recent literature emphasizing pain management is extensive and widely available for all anesthesiologists. In addition, residents rotated between the two hospitals, and over the two-year study period several of these residents spent time at both the study and control hospitals. Therefore, discussion of pain management strategies with the residents could have influenced practice at the control hospital. Policies for drug availability and distribution, as well as postoperative nursing care, were not uniform across the two hospitals but may have facilitated the use of PCA at the control hospital while restricting an increased use of other promoted strategies.
For the specialty of anesthesia no similar studies of changing practice patterns and outcome are available. Studies which have examined physician behavior have found that changing physician practice is a complex issue . Lomas  noted that to change clinical practice involved a series of educational, administrative, and economic incentives or disincentives. Using this methodology, only a few studies have been successful in achieving a practice change and improved patient outcome. Our study avoided rigid patient protocols, practice guidelines, and economic incentives and allowed the anesthesiologists to choose the promoted strategies based only on the educational and feedback interventions. The avoidance of these additional approaches may have limited our ability to affect changes in practice and outcome.
There were several limitations to the study related to the directed interventions. Perhaps if more opinion leaders had been designated, the daily promotion of the strategies could have been even more successful. Several anesthesiologists questioned the strategies promoted during the educational period. They were reluctant to use new strategies when the efficacies were not fully documented and there was a potential for increased patient risk. It is possible that some anesthesiologists may not have read the distributed articles. Several did not attend the educational seminars due to scheduling conflicts and holidays. Some expressed a dislike of the feedback forms. Criticisms included the relatively small number of patients reported for some anesthesiologists, inadequate case mix adjustment, and the choice of the comparator group (the three anesthesiologists whose patients had the lowest pain scores). Although the feedback forms were designed with simple tables and graphs, some anesthesiologists did not understand their personalized report. One intervention not used, one-on-one interviews with anesthesiologists, might have been helpful to understand the barriers to change as there were a few anesthesiologists who made absolutely no change in their practice of pain management.
Were the outcome measures we used adequate to demonstrate a benefit of the promoted strategies? Postoperative interviews were restricted to two questions concerning pain-pain at rest and pain with activity. A retrospective pain score assessed the entire interval for six hours after PACU discharge. This may be one of the most difficult time periods to assess and successfully treat postoperative pain. Some patients may still be drowsy after anesthesia which could have distorted their evaluation of pain during this time period and restricted their use of PCA.
Although this study showed no indication of increased use of resources (hospital and PACU lengths of stay over the study period), the increased costs (capital and operating) required for the promoted interventions were not computed. Similarly, the determination of costs for nursing and patient education was not a part of the study protocol.
In conclusion, directed interventions aimed at anesthesiologists changed practice patterns. After educational initiatives and personalized feedback, the proportion of patients who received PCA and NSAIDs increased at the study hospital compared to a control hospital with no directed interventions. However, it was more difficult to demonstrate an effect on patient outcomes; only a small decrease in early postoperative pain with activity was found.
The authors wish to thank Ms. Mary Tan, Ms. Nadine Lyle, and Ms. Kathy Sykora for data collection, preparation, and statistical analysis.
Complications in the PACU
1. Excessive pain: Moaning or writhing in pain at any time in the PACU or initial patient care dominated by pain control
2. Critical respiratory event: Hypoxemia (hemoglobin oxygen saturation <90%), hypoventilation (respiratory rate <8 breaths/min or arterial CO2 tension >50 mm Hg), or upper airway obstruction (stridor or laryngospasm) requiring any active or specific intervention (ventilation, tracheal intubation, opioid or muscle relaxant antagonism, insertion of oral/nasal airway, or airway manipulation)
3. Room air desaturation: Hemoglobin oxygen saturation <90% occurring after removal of oxygen therapy in the PACU
Problems in the Early Postoperative Interval (the first 6 h after PACU discharge)
1. Postoperative pain: Measured at rest and with activity (movement or deep breaths) using a visual analog scale (0-10; 0 indicating no pain and 10 the worst pain ever)
2. Naloxone requirement: Any dose of Naloxone documented on the patient record in the first 6 h after PACU discharge
3. Urinary retention: Any patient complaint or documentation of bladder distention requiring catheterization
1. Cooper J, Cullen D, Nemeskai R, et al. Effects of information feedback and pulse oximetry on the incidence of anesthesia complications. Anesthesiology 1987;67:686-94.
2. Yeager M, Glass D, Neff R, Brinck-Johnsen T. Epidural anaesthesia and analgesia in high-risk surgical patients. Anesthesiology 1987;66:729-36.
3. Engberg G, Wiklund L. Pulmonary complications after upper abdominal surgery: their prevention with intercostal blocks. Acta Anaesthesiol Scand 1988;32:1-9.
4. Soumerai SB, Avorn J. Principles of educational outreach ("academic detailing") to improve clinical decision making. JAMA 1990;263:549-56.
5. Davis DA, Thomson MA, Oxman AD, Haynes RB. Changing physician performance. A systematic review of the effect of continuing medical education strategies. JAMA 1995;274:700-5.
6. Task Force on Acute Pain. Adult postoperative pain. In: Ready BL, Edwards WT, eds. Management of acute pain: a practical guide. Seattle, WA: IASP Publications, 1992:22-5.
7. Ferrante FM. Acute pain management. International Anesthesia Research Society 1993 Review Course Lectures, 1993:102-8.
8. Moote C. Techniques for post-op pain management in the adult. Can J Anaesth 1993;40:R19-24.
9. Gould TH, Crosby DL, Harmer M, et al. Policy for controlling pain after surgery: effect of sequential changes in management. BMJ 1992;305:1187-93.
10. Ready LB. Patient-controlled analgesia-does it provide more than comfort? Can J Anaesth 1990;37:719-21.
11. Laitinen J, Nuutinen L. Intravenous diclofenac coupled with PCA fentanyl for pain relief after total hip replacement. Anesthesiology 1992;76:194-8.
12. McGlew IC, Angliss DB, Gee GJ, et al. A comparison of rectal indomethacin with placebo for pain relief following spinal surgery. Anaesth Intensive Care 1991;19:40-5.
13. Segstro R, Morley-Forster PK, Lu G. Indomethacin as a postoperative analgesic for total hip arthroplasty. Can J Anaesth 1991;38:578-81.
14. Zikowski D, Hord AH, Haddox JD, Glascock J. Ketorolac-induced bronchospasm. Anesth Analg 1993;76:417-9.
15. Haddow GR, Riley E, Isaacs R, McSharry R. Ketorolac, nasal polyposis, and bronchial asthma: a cause for concern. Anesth Analg 1993;76:420-2.
16. Gillberg LE, Harsten AS, Stahl BL. Preoperative diclofenac sodium reduces post-laparoscopy pain. Can J Anaesth 1993;40:406-8.
17. Kinsella J, Moffat AC, Patrick JW, et al. Ketorolac trometamol for postoperative analgesia after orthopaedic surgery. Br J Anaesth 1992;69:19-22.
18. Parker RK, Holtmann B, Smith I, White PF. Use of ketorolac after lower abdominal surgery. Effect on analgesic requirement and surgical outcome. Anesthesiology 1994;80:6-12.
19. Beattie WS, Buckley DN, Forrest JB. Epidural morphine reduces the risk of postoperative myocardial ischaemia in patients with cardiac risk factors. Can J Anaesth 1993;40:532-41.
20. Liu WHD, Aitkenhead AR. Comparison of contemporaneous and retrospective assessment of postoperative pain using the visual analogue scale. Br J Anaesth 1991;67:768-71.
21. Fleiss JL. Statistical methods for rates and proportions. New York: John Wiley & Sons, 1981.
22. Mantha S, Thisted R, Foss J, et al. A proposal to use confidence intervals for visual analog scale data for pain measurement to determine clinical significance. Anesth Analg 1993;77:1041-7.
23. Choiniere M, Grenier R, Paquette C. Patient-controlled analgesia: a double-blind study in burn patients. Anaesthesia 1992;47:467-72.
24. Dixon AS. The evolution of clinical policies. Med Care 1990;28:201-20.
© 1997 International Anesthesia Research Society
25. Lomas J. Promoting clinical policy change: using the art to promote the science in medicine. In: Andersen TF, Mooney G. The challenges of medical practice variations. London: MacMillan, 1990:174-91.