Patient-controlled analgesia (PCA) offers significant advantages over traditional on-demand injection of opioids for postoperative pain control [1,2]. These include the allowance for the variability among patients, its simplicity, and the inherent safety factors . PCA has been associated with improved pain control, fewer side effects, and shorter hospital stay [3,4]. PCA is routinely available in most hospitals in the United States, but PCA management practices differ markedly among states and hospitals . In some settings, postoperative PCA prescriptions and decisions are handled by a dedicated acute pain service (APS), whereas in others, PCA is managed by primary service (PS) physicians (most often surgeons providing postoperative care). Although an implicit assumption of pain clinicians is that a dedicated APS provides superior service and increased patient benefits, no study has systematically compared APS with PS PCA management.
At the University of Pittsburgh Medical Center (UPMC), we were presented with a unique opportunity in 1991. Before that time, the UPMC APS had been responsible for all PCA prescriptions and all PCA management. After a reevaluation of the APS, a general consensus among surgeons and APS physicians emerged. It was decided that routine postoperative PCA management could be provided by the surgical staff, with the APS being consulted at the discretion of the surgical staff for more complicated patients (e.g., patients who either cannot be made comfortable by the PS's adjustment of the PCA or are consuming a large amount of opioid). After discussions with the Chairman of Surgery, who gave permission for prospective data to be collected without the knowledge of the individual surgeons writing the PCA prescriptions, a comparative study was conducted. The overall objective of this study was to evaluate differences in PCA management between PS physicians and APS physicians and to evaluate the associated patient outcomes for pain relief and incidence of side effects. More specifically, we hypothesized that patients treated by APS physicians would have fewer side effects, lower pain scores, more frequent PCA dose adjustments, and a greater likelihood of receiving a loading dose compared with patients treated by PS physicians.
After approval of the protocol by the institutional review board, 40 consecutive patients prescribed PCA by a PS physician were studied (PS group). The data were collected prospectively over a 3-mo period 1 yr after the PS assumed responsibility for PCA management. All PS group subjects were inpatients on UPMC surgical services who were followed by surgical house staff and attending surgeons. The nurses and physicians directly responsible for care were unaware that data were being collected regarding PCA management or postoperative pain levels.
PS group data were collected daily by a research nurse and included patient demographics, PCA prescription, changes in PCA orders, opioid consumed, reasons for terminating PCA, verbal analog scale (VAS) pain scores, side effects (nausea or emesis, sedation, respiratory depression, and mental status changes were determined every 4 h while on PCA), and post-PCA pain management information. VAS scores were obtained by the nursing staff as part of the routine postoperative PCA orders.
Prior to the data collection period, an informational pamphlet on PCA management was distributed to all house staff at UPMC, guidelines for PCA prescription were placed on all nursing units, educational sessions on PCA were made available to the surgical services, and the nursing staff received further education about PCA.
After the collection of PS data, an APS comparison group of 40 PCA-treatment patients was randomly selected from an existing pool of patients previously treated by the APS the year before PS took over PCA. A proportionate stratified random sampling procedure was used to select APS cases to match the age, gender, and type of surgery characteristics of the PS group. Data were collected from patients' medical record and an APS database. The VAS scores for the APS group were obtained by the APS and the nursing staff.
The APS consists of a full-time nurse coordinator, anesthesiology house staff, pain management fellows, and anesthesiology pain management attendings. All of the attending pain physicians had completed formal pain fellowship training. Neither the nursing staff nor the APS members were aware that the data were to be used for comparative purposes.
Group differences on continuous data were evaluated by general linear model one-way analysis of variance (ANOVA), using the SYSTAT software package (version 5; SYSTAT Inc., Evanston, IL) . Nonparametric categorical and dichotomous outcome measures were tested by chi squared procedures, using the StatXact software program (CYTEL Software Corp., Cambridge, MA) to compute exact probabilities for these analyses . A P value of less than 0.05 was considered statistically significant.
One case from the PS group and one from the APS group were eliminated from further analysis due to insufficient data on the primary outcome measures. The demographic data for the remaining 39 patients in each group are summarized in Table 1. There were no significant differences between the PS and APS patient groups with respect to age, gender, weight, ASA physical status, and type of surgery. These findings were expected and indicate that the stratified sampling procedures successfully matched the APS patients selected to the PS group characteristics.
(Table 2) presents the results for the primary outcome measures. APS physicians maintained patients on PCA significantly longer than PS physicians. APS physicians also ordered loading doses, adjusted the PCA dosage, utilized opioids other than morphine, employed continuous infusions, and prescribed per os medications (including per os morphine) significantly more often than PS physicians. As a result of these differences, APS-treated patients consumed more opioid in milligrams of morphine equivalents, both in terms of the total amount and the average daily amount. On the other hand, PS physicians ordered intramuscular (IM) medications significantly more often than APS physicians (Table 2).
As displayed in Table 2, VAS pain scores between the APS and PS groups were not significantly different on PCA Day 1 or 2 or when PCA was discontinued. To understand the relatively higher pain scores during the first two days of PCA treatment, point-biserial correlations were conducted between patients' history of previous use of opioids and their daily pain scores. These correlations were significant (all r values >0.53, P < 0.001). Patients with a history of previous opioid usage scored an average of 2.8 points higher on the 0-10 VAS pain rating scale compared with patients who did not have a history of opioid use. Thus, since more than 25% of both the PS and APS sample had a history of previous opioid usage, the mean Days 1 and 2 pain scores were inflated significantly by the inclusion of these patients. The strength of the association between opioid history and pain scores at discontinuation was less than that observed for the first two days of PCA but remained statistically significant (r = 0.41, P < 0.01). Patients without a history of previous opioid usage had discontinuation pain scores 1.8 point less on average than patients reporting previous opioid usage.
Compared with APS patients, PS patients were found to have a significantly higher incidence rate of side effects (Table 3). The PS patients had significantly higher rates of urinary retention and nausea when compared with the APS patients (Table 3). The PS and APS groups were not significantly different with respect to pruritus rates, and no patient experienced respiratory depression (Table 3).
Intravenous PCA is often described as self-explanatory, simple, and efficacious. Nevertheless, this study suggests that a dedicated APS uses PCA technology differently than non-APS physicians. Specifically, if patients reported previous adverse experiences with morphine, it was avoided by the APS. The APS more often provided loading doses, adjusted the dose, or added continuous infusion, resulting in increased opioid use. When PCA was discontinued, APS patients were more likely to be converted to oral medications, often utilizing oral morphine and a tapering schedule, thereby avoiding the intramuscular route. In conjunction with daily interactions focused on treating pain and side effects, these practice differences presumably allowed the APS patients to experience fewer side effects, particularly with respect to nausea and urinary retention.
Subjects in both groups reported pain scores in the range 3.7-5.1 on a 0-10 VAS. PCA studies often report lower pain scores, but given the design of our study, scores in this range are understandable. First, the pain scores represent the patients' average pain score, incorporating both movement and rest scores. Movement scores less than 5 are acceptable in other PCA studies  and in general, average postoperative pain scores often range between 2 and 5 [8-10]. Second, patient populations known to have increased pain scores and increased opioid consumption, such as patients with chronic pain, chronic opioid consumption, past or present substance abuse, and neuropathic pain [11-15], were not excluded from eligibility. These conditions are common exclusion criteria in many recent PCA studies [9,10,16]. In fact, more than 25% of our subjects had a history of preoperative opioid use, which we demonstrated contributed to our higher pain scores (see Table 2). We wanted to compare a broad spectrum of PCA patients; thus, we included such patients in our study.
The potential sources of the practice differences are many. The standard UPMC APS approach to the postoperative patient using PCA includes a patient visit every morning. Before each visit, the chart is reviewed, and any new medication orders or changes in status are noted. The patient's opioid consumption and pattern of use in the previous 24 hours is determined. The patient is interviewed regarding pain relief, satisfaction with PCA, quality of sleep, and side effects. The goals of pain management (minimum pain and side effects) and the availability of treatment for side effects are reinforced daily. The PCA settings and syringe contents are confirmed, and the prescription is adjusted based on a variety of factors, including quality of pain relief, opioid requirements, sleep, activity levels, and side effects. Patients with difficult pain control are visited again in the afternoon. When the oral route is established, PCA is discontinued, and equivalent oral analgesics are provided in a tapering dose.
This APS approach requires a brief, intense interaction with the patient focused on achieving the best possible analgesia. When PS physicians visit their postoperative patients, they have many goals, including the assessment of overall physiologic status (renal, pulmonary, cardiovascular, etc.), wound healing, progress in resuming normal activities (ambulation, diet, bowel function), and pain relief and patient comfort, as well as the prevention, identification, and treatment of complications. Pain relief is but one of many important areas of concern and, therefore, is unlikely to be evaluated as thoroughly by a PS physician as by a dedicated APS physician. Evidence to support this contention was found in this study. The PS daily progress notes in patient charts contained no specific information about PCA dose or opioid consumption but did have frequent comments about adequacy of analgesia and the "presence" of PCA.
Our results confirm the observation made by Smythe et al.  that common reasons for discontinuation of PCA are inadequate pain control and/or the presence of side effects. It is known that wide variability in postoperative pain experiences and opioid requirements are common, but in this study, the PS subjects infrequently received changes in their PCA prescription for complaints about side effects or inadequate analgesia. Instead, it appears that complaints led to the abandonment of PCA for alternative therapy, such as IM injections. This suggests that the variability of opioid responsiveness and the factors associated with such variability are not adequately appreciated. There may be a tendency to believe that PCA didn't work for this patient rather than to change the parameters of the PCA. Recent quality improvement data from UPMC verify that orthopedic surgery patients whose postoperative pain was managed by intravenous or IM injections frequently had "failed" PCA. These findings reinforce Bonica's  observation that "inadequate or improper application of available information and therapies is certainly the most important reason for inadequate postoperative pain relief."
Concerns about the costs of PCA , reimbursement for PCA , or costs of APS consultation [21,22] may be advanced as reasons either to avoid PCA altogether or to make PCA a PS-provided service. Although medication and supply costs are often greater for PCA than for IM injections, savings may accrue from decreased nursing time and costs, lower complication rates, or shortened hospital stay [3,23-25]. Thus, the investment in PCA technology may pay off in overall cost-savings. In terms of the costs of an APS, although additional patient charges may be incurred, the benefits-whether monetary or in quality of care-simply are not known. No systematic comparison of APS versus PS management of postoperative patients has been performed. Such a study may be difficult to control adequately, as physicians' awareness of being analyzed is likely to influence the care provided. Given these potential limitations of a randomized comparative study, the design of this study allowed for an observational, "real-world" comparison representing unbiased, true practice parameters.
The findings provide evidence that an APS staffed by physicians and nurses specializing in the assessment and management of pain results in significant practice differences compared with PS management. Additional studies are needed to determine whether these differences lead to differences in patient satisfaction or whether PCA management can be improved by further education of staff or modification of the parameters of PCA care.
1. White PF. Use of patient-controlled analgesia for management of acute pain. JAMA 1988;259:243-7.
2. Ballantyne JC, Card DB, Chalmers TC, et al. Postoperative patient-controlled analgesia: meta-analysis of initial randomized control trials. J Clin Anesth 1993;5:181-92.
3. Wasylak TJ, Abbott FV, English MJM, Jeans ME. Reduction of post-operative morbidity following patient-controlled morphine. Can J Anaesth 1990;37:726-31.
4. Thomas V, Heath M, Rose D, Flory P. Psychological characteristics and the effectiveness of patient-controlled analgesia. Br J Anaesth 1995;74:271-6.
5. Ready LB. How many acute pain services are there in the United States, and who is managing patient-controlled analgesia? [letter]. Anesthesiology 1995;82:322.
6. Wilkinson L. SYSTAT for Windows, version 5. Evanston, IL: SYSTAT Inc., 1992.
7. Mehta CR, Patel NR, Gray R. On computing an exact confidence interval for the common odds ratio in several 2 x 2 contingency tables. J Am Stat Assoc 1985;80:969-73.
8. De Leon-Casasola, Lema MJ, Karabella D, Harrison P. Postoperative myocardial ischemia: epidural versus intravenous patient-controlled analgesia-a pilot project. Reg Anesth 1995;20:105-12.
9. Mansfield MD, James KS, Kinsella J. Influence of dose and timing of administration of morphine on postoperative pain and analgesic requirements. Br J Anaesth 1996;76:358-61.
10. Van Lancker P, Devulder J, Rolly G. Systemic piroxicam as an adjunct to patient-controlled analgesia with alfentanil for postoperative pain relief. Anaesthesia 1996;51:658-62.
11. Jadad AR, Carroll D, Glynn CJ, et al. Morphine responsiveness of chronic pain: double-blind randomised crossover study with patient-controlled analgesia. Lancet 1992;339:1367-71.
12. Rapp SE, Ready LB, Nessly ML. Acute pain management in patients with prior opioid consumption: a case-controlled retrospective review. Pain 1995;61:196-201.
13. Preble LM, Guveyan JA, Sinatra RS. Patient characteristics influencing postoperative pain management. In: Sinatra RS, Hord AH, Ginsberg B, Preble LM, eds. Acute pain: mechanisms and management. St. Louis, MO: Mosby-Year Book, 1992:140-50.
14. Burns JW, Hodsman NBA, McLintock TTC, et al. The influence of patient characteristics on the requirements for postoperative analgesia: a reassessment using patient-controlled analgesia. Anaesthesia 1989;44:2-6.
15. Owen H, White PF. Patient-controlled analgesia: an overview. In: Sinatra RS, Hord AH, Ginsberg B, Preble LM, eds. Acute pain: mechanisms and management. St. Louis, MO: Mosby-Year Book, 1992:151-64.
16. Rogers JE, Fleming BG, Macintosh KC, et al. Effect of timing of ketorolac administration on patient-controlled opioid use. Br J Anaesth 1995;75:15-8.
17. Smythe MA, Martin SE, Schad RF. Retrospective evaluation of patient-controlled analgesia use in a large teaching institution. DICP 1991;25:1058-61.
18. Bonica JJ. Postoperative pain. In: Bonica JJ, ed. The management of pain. 2nd Ed. Philadelphia: Lea & Febiger, 1990:461-80.
19. Moote CA. Postoperative pain management-back to basics [editorial]. Can J Anaesth 1995;42:453-7.
20. Mackey DC, Ebener MK, Howe BL. Patient-controlled analgesia and the acute pain service in the United States: Health-Care Financing Administration policy is impeding optimal patientcontrolled analgesia management [letter]. Anesthesiology 1995;83:433-4.
21. Moote CA. Pharmacoeconomics of patient-controlled analgesia. Perspect Pain Manag 1994;3:7-12.
22. Rawal N, Berggren L. Organization of acute pain services: a low-cost model. Pain 1994;57:117-23.
23. Chan VWS, Chung F, McQuestion M, Gomez M. Impact of patient-controlled analgesia on required nursing time and duration of postoperative recovery. Reg Anesth 1995;20:506-14.
24. Ready LB. Patient-controlled analgesia-does it provide more than comfort? [editorial]. Can J Anaesth 1990;37:719-21.
25. Koh P, Thomas VJ. Patient-controlled analgesia (PCA): does time saved by PCA improve patient satisfaction with nursing care? [review]. J Adv Nurs 1994;20:61-70.