PAIN Evaluation of costs and effects of epidural analgesia and patient-controlled intravenous analgesia after major abdominal surgery

Transcription

1 British Journal of Anaesthesia 96 (1): (2005) doi: /bja/aei270 Advance Access publication October 28, 2005 PAIN Evaluation of costs and effects of epidural analgesia and patient-controlled intravenous analgesia after major abdominal surgery E. Bartha 1 *, P. Carlsson 2 and S. Kalman 1 1 Department of Anaesthesiology, Linköping University Hospital. 2 Centre for Medical Technology Assessment, Linköping University, Linköping, Sweden *Corresponding author. elisabeth.bartha@lio.se Background. The outcome of different treatment strategies for postoperative pain has been an issue of controversy. Apart from efficacy and effectiveness a policy decision should also consider cost-effectiveness. Since economic analyses on postoperative pain treatment are rare we developed a decision model in a pilot cost-effectiveness analysis (CEA) comparing epidural analgesia (EDA) and patient-controlled intravenous analgesia (PCIA) after major abdominal surgery in routine care. Methods. Using a decision-tree model, treatment with EDA (ropivacaine and morphine) was compared with PCIA (morphine). Effects and costs of treatment were established. The number of pain-free days at rest (pain intensity <30 using visual analogue scale mm) was the primary measure of effect. An incremental cost-effectiveness ratio (ICER) was calculated as the difference in direct costs divided by the difference in effect. A database on 644 patients collected for the purpose of quality control during the period of 1997 to 1999 was the main data source. Sensitivity analysis was used to test uncertain data. Results. EDA was more effective in terms of pain-free days but more expensive. The additional cost for each pain-free day was 5652 Euros. Conclusion. It is a judgement of value if the additional cost is reasonable. When the cost of around Euros per gained life-year with full health for other interventions is debated, our result indicates poor cost-effectiveness for EDA. Before any conclusion can be drawn concerning policy recommendations the difference in costs has to be related to other outcome measures as length of hospital stay, morbidity and mortality are required. Br J Anaesth 2006; 96: Keywords: analgesia, postoperative; analgesia, patient-controlled; analgesic techniques, extradural; cost-effectiveness Accepted for publication: October 3, 2005 Good postoperative pain control is a mandatory component of adequate postoperative care if accelerated recovery is aimed for. 1 3 The most common pain relief strategies after major abdominal surgery are epidural infusion of local anaesthetics and opiates (EDA) and/or patientcontrolled administration of intravenous opiates (PCIA). According to the guidelines of the Swedish Society of Anaesthesiology both methods are accepted options following major surgery. In general when selecting different treatment strategies efficacy, effectiveness and costs should be taken into account. At the University Hospital in Linköping (Sweden) both EDA and PCIA have been used following major abdominal surgery. Assessment of the clinical effectiveness of EDA during the period of 1997 to 1999 revealed, in line with the findings of others, 4 7 that 10% of patients scheduled for epidural analgesia had their treatment discontinued because of technical problems, minor side-effects or insufficient pain relief. Hence costs and consequences of planned and discontinued treatment became of interest when comparing these two strategies. Moreover this decision-problem concerns a lot of patients. In the county of Östergötland in Sweden with Ó The Board of Management and Trustees of the British Journal of Anaesthesia All rights reserved. For Permissions, please journals.permissions@oxfordjournals.org

2 Bartha et al inhabitants, 930 postoperative patients received epidural analgesia and 800 PCIA during This would imply that in Sweden patients are treated with either EDA or PCIA every year. Since there are few economic analyses of postoperative pain treatment a decision analytic model was developed 8 to estimate the cost-effectiveness of epidural analgesia compared with patient-controlled intravenous analgesia (PCIA) after major abdominal surgery under ordinary clinical circumstances. Methods Cost-effectiveness model A decision tree was used to model the clinical pathways for estimating the effects (E) and costs (C) of treatment with EDA and PCIA. The measure of effect was expressed as number of pain-free days. The cost-effectiveness, the average cost for reaching a particular outcome for a given treatment, is expressed as cost-effectiveness ratio (CER) (Fig. 1). CER¼ Cost Effect : When a decision has to be made to replace a treatment with a more expensive and more effective treatment, an estimate of the additional resources that have to be used to obtain the additional benefit is needed. 9 That is the incremental cost-effectiveness ratio (ICER): the difference in direct cost (DC) divided by the difference in effect (DE) between the two alternatives (Fig. 1). ICER¼ ðcost EDA Cost PCIA Þ ðeffect EDA Effect PCIA Þ : The result is presented both as CER and as ICER (Fig. 1). All costs are in 2005 price and are converted to Euros using Cost EDA EDA the exchange rate: 1 Euro¼9 Swedish crowns. Sensitivity analyses were performed to estimate the cost-effectiveness by testing data that were uncertain or debatable. Data source The main data source was an existing local database, started in 1997 for assessment of clinical effectiveness and by 1999 included 644 consecutive patients treated with EDA (n¼602) or PCIA (n¼42) following major abdominal surgery (Table 1). Patients were selected for treatment with PCIA in a non-systematic way if a shortage of resources existed on the PCU/ICU or if the patient refused epidural analgesia. The data of the 42 patients treated with PCIA and of the 569 patients treated with EDA were used, but the data of 33 patients were incomplete. The Research Ethics Committee of the University Hospital in Linköping approved the use of the database as data source for analysis. The treatment strategies For the thoracic epidural analgesia a mixture of ropivacaine 2mgml 1 with morphine 0.03 mg ml 1 was delivered at a constant infusion rate of 5.5 ml h 1 (Baxter Infuser LV, Baxter Healthcare Corporation, Deerfield, IL). Rescue pain treatment was given according to an algorithm. The patients were observed on the Postoperative Care Unit/ Intensive Care Unit (PCU/ICU) for at least 12 h according to actual guidelines. For treatment with PCIA, morphine 5 mg ml 1 was used, and the pump (Smiths Medical, Deltec, Inc., St Paul, MN, USA) was programmed individually to meet requirements and delivery was on demand. The duration of the care on the PCU/ICU was shorter compared with the EDA group according to local instructions. The duration of the treatment was 3 days for both patient groups. Estimation of probabilities The structure of the decision tree was developed by the selection of the relevant events and pathways: complete treatment and change of treatment strategy; unsuccessful CER EDA ICER Table 1 Patient characteristics of the epidural and the PCIA group. Age and weight are given in mean values (SD). The others are presented in percent EDA total (n¼569) PCIA (n¼42) Cost PCIA CER PCIA PCIA Effect PCIA Effect EDA Fig 1 Illustration of the cost-effectiveness ratio (CER) and the incremental cost-effectiveness ratio (ICER). 10 The slopes of the lines from origin give the CER for the treatment with EDA and PCIA. The ICER (dotted line) is the slope of the line joining the points EDA and PCIA. CER¼Cost/Effect and ICER¼(Cost EDA Cost PCIA )/(Effect EDA Effect PCIA ). Mean age, years (SD) 57 (17) 48 (16) Weight, kg (SD) 72 (15) 77 (22) Female, % Male, % ASA I, % ASA II, % ASA III, % ASA IV, % 2 0 High laparotomy, % Low laparotomy, % Urology, % 8 14 Trauma, %

3 Economic evaluation of postoperative epidural and patient-controlled intravenous analgesia Table 2 The selected courses of treatment and the actual probabilities. Data are obtained from the database for the epidural and for the PCIA group. *Low blood pressure, pruritus and respiratory depression EDA (n¼569) PCIA (n¼42) Unsuccessful attempt, treatment with PCIA n.a Poor effect, reinsertion of epidural catheter n.a Successful attempt and start of EDA n.a Poor effect, side-effect* and change of 0.09 n.a treatment to PCIA Poor effect, side-effect* and change of treatment to bolus doses of morphine Minor problems with the epidural catheter, motor blockade and numbness, continued treatment Accidental dislocation of epidural catheter 0.06 n.a Fulfilled treatment (EDA, or PCIA) Pain during treatment, further intervention No pain during treatment 0.7 n.a Nausea and vomiting (with and without pharmacological treatment) Respiratory depression (no pharmacological treatment was needed) attempt to introduce epidural catheter; early dislocation of catheter; reinsertion of epidural catheter and need for additional pain treatment. The decision tree was supplemented with probabilities for different pathways using data from the database (Table 2 and Fig. 2). The structure of the model was simplified by one assumption: the costs and the number of pain-free days had a linear relationship. Estimation of the effectiveness The pain intensity was monitored every 3 h using a visual analogue scale both at rest and during activity (VAS mm) and a representative value for each day was recorded in the database. The effectiveness of treatment for the cost-effectiveness analysis was expressed as the number of pain-free days at rest, as is shown in the decision tree. A pain-free day was defined as having a VAS<30 a whole day. The analgesic effectiveness was 3 if the patient had 3 pain-free days, 2 in the case of 2 pain-free days, 1 in the case of 1 pain-free day, and 0 if the patient did not experience VAS<30 at all (Fig. 2). The number of pain-free days was weighted by the overall probabilities for each branch of the tree, and the sum of these resulted in the expected number of pain-free days for each treatment strategy (Fig. 2). The same procedure was done using probabilities at rest and at activity. Estimation of the costs The most relevant, direct medical costs were evaluated: the cost of human resources and also the costs of medical devices and drugs. The time used for the selected activities cannot be used for other simultaneous activities and hence represents a cost. Data on cost per hour for the staff and data on devices and drugs were obtained from the hospital administration and from the hospital pharmacy (Table 3). The time devoted to certain activities was measured or based on interviews with the staff. The cost of particular interventions related to problems with the epidural catheter at the insertion site (leakage, bleeding) or to minor side-effects (nausea, vomiting) was weighted based on the probabilities of these events. All calculated costs were based on 3-day long treatment periods. The cost of postoperative care was derived from official tariffs. The identified costs of the clinical pathways during treatment were weighted by the actual probabilities demonstrated on the decision tree. Sensitivity analysis Sensitivity analysis was performed in order to test the robustness of the result. One approach was to test the most uncertain cost item in the model, the tariff, for postoperative care, as it was unclear how it was set in relation to real costs of resources used on the PCU/ICU. We have tested the model using the lowest reported tariff assuming that the EDA and the PCIA group had the same length (3 4 h) of care on the ICU/PCU. Another uncertainty was the effect, as it is a matter of dispute if the most valid measure of pain is estimated at rest or during activity. Therefore the model was tested by using the higher incremental effect of EDA over PCIA during activity. A second approach was the scenario analysis, when one optimal scenario of treatment was tested by reducing the probabilities of the technical difficulties related to EDA: catheter dislocation, unsuccessful attempt, reinsertion of catheter and the need for additional pain treatment (Table 4). At last we tested the influence of potential bias related to the data derived from a non-randomized population. Each patient treated with PCIA in the database was matched in pairs with one patient treated with EDA. First, the age (it was allowed to vary by 2 yr) and the type of surgical intervention and second, the gender and ASA-group were used in the matching procedure. Data of the matched subgroup of epidural patients was used as described in methods, and the ICER was also calculated. Results Analysis of the effectiveness The number of the expected pain-free days at rest for a patient in the model treated with EDA was 2.36 and with PCIA 2.17, yielding an incremental effect of 0.19 day (E¼Effect EDA Effect PCIA ). During activity the number of expected pain-free days was lower: 1.86 in the EDA and 1.27 in the PCIA group and the incremental effect was more than a half day, 0.59 (Table 5). The overall probabilities for different clinical pathways during treatment with EDA or PCIA can also be calculated using the decision tree (Fig. 2). For example, the probability of achieving 3 pain-free days without any additional pain treatment and without any 113

4 Bartha et al. Decision node Pain-free 3 days Chance node Pain-free 2 days End node Fulfilled treatment Pain-free 1 day Never pain-free PCIA (n=42) Pain-free 3 days Pain-free 2 days Change of treatment Pain-free 1 day Never pain-free Operated patients (n=611) Pain-free 3 days Pain-free 2 days Unsuccessful attempt start PCIA Pain-free 1 day Never pain-free Pain-free 3 days Pain-free 2 days Reinsertion of EDA catheter, fulfilled treatment Pain-free 1 day Never pain-free Pain-free 3 days Pain-free 2 days Poor effect, side-effects, change to PCIA Pain-free 1 day EDA (n=569) Never pain-free Pain-free 3 days Pain-free 2 days Poor effect, side-effects, change to i.v. morphine Pain-free 1 day Successful start of epidural analgesia Never pain-free Pain-free 3 days Pain-free 2 days Accidental dislocation of catheter Pain-free 1 day Never pain-free Pain-free 3 days Need for intervention Pain-free 2 days Pain-free 1 day Continued EDA Never pain-free Pain-free 3 days Pain-free 2 days No need for intervention Pain-free 1 day Never pain-free Fig 2 The decision tree illustrates the selected clinical pathways and probabilities for treatment with EDA and PCIA. The final branches show the effectiveness of treatment expressed as number of pain-free days with actual probabilities at rest and during activity. Finally the overall probability (product of the probabilities) for the actual branch is given using probabilities only at rest because of the limited space available. 114

5 Economic evaluation of postoperative epidural and patient-controlled intravenous analgesia Table 3 The calculated costs for staff and pharmacy charges, costs for material and drugs. The costs are calculated for each patient and for a 3-day treatment period Costs for personnel, material and drugs EDA (Euro) PCA (Euro) Personnel costs for starting the EDA Personnel costs for unsuccessful attempt to introduce EDA Personnel costs on the general ward Personnel costs for intervention for pain relief if needed Personnel costs for changing treatment strategy (side-effects) Personnel cost for minor side-effects and 13 0 technical problems with the epidural catheter Personnel costs for intermittent injection of morphine as required (6 injections) Pharmacy charges Cost for material 12 Included above Tariff for postoperative care Table 4 Probabilities used for optimal scenario. Both the probabilities from the database (base case) and the assumed probabilities (optimal scenario) for catheter dislocation, unsuccessful attempt, reinsertion of catheter and for change of treatment because of side effects or poor pain relief are given Probabilities sampled from the database: Base case Optimal scenario Unsuccessful attempt, reinsertion of EDA catheter Early dislocation of EDA catheter Poor pain relief, side-effects, change of treatment strategy Table 5 The expected number of pain-free days, (rest and activity) and the costs of postoperative epidural analgesia and patient-controlled intravenous morphine analgesia. The incremental effect and cost is also demonstrated. The costeffectiveness ratio (CER) and the incremental cost-effectiveness ratio (ICER) are calculated using the measure of effect at rest (base case result). CER¼Cost/ Effect and ICER¼(Cost EDA Cost PCIA )/(Effect EDA Effect PCIA ) Cost/patient (Euros) Effect Number of pain-free days Rest Activity CER Cost/pain-free day/patient (Euros) EDA PCIA Increment of EDA to PCIA (ICER) problem with the epidural catheter was 0.49 at rest and 0.41 during activity with treatment with EDA. The corresponding values for the PCIA group were 0.56 at rest and 0.28 during activity. Cost analysis, calculation of the expected cost The expected cost per patient in the EDA group was 1701 Euros and in the PCIA group 627 Euros. The difference in costs (C¼Cost EDA Cost PCIA ) was 1074 Euros (Table 5). Table 6 Sensitivity analysis was performed to estimate the influence of uncertain items on cost and ICER. The model was tested by the lowest tarrif for PCU/ICU, by the higher incremental effect of EDA over PCIA at activity, by an optimized scenario and by a matching procedure. The used probabilities for the scenario analysis are demonstrated in Table 4 Base case Lowest tariff for care on PCU/ICU Pain at activity Scenario analysis Matching procedure Cost/patient for EDA (Euro) Number of pain-free days E¼Effect EDA Effect PCIA ICER The cost-effectiveness ratio (CER), incremental cost-effectiveness ratio (ICER) The cost per pain-free day at rest was 721 Euros per patient for treatment with EDA and 289 Euros per patient for treatment with PCIA. The incremental cost-effectiveness ratio (ICER) at rest base case result was 5625 Euros (Table 5). Sensitivity analysis Using the lowest tariff, assuming the same length of care on the ICU/PCU, the base case ICER decreased by 66% from 5625 Euros to 1448 Euros (Table 6). When the influence of better pain relief of EDA was tested by using the measure of effect during activity, the ICER decreased by 67% from 5625 Euros to 1896 Euros (Table 6). The scenario analysis, when the probabilities in the model affected by the technical difficulties were optimized (Table 4), the cost of EDA decreased only by 2% to 1666 Euros; in contrast the incremental effect increased by 105% to 0.39 days which resulted in a decrease of ICER by 52% to 2664 Euros. The calculated ICER using data of the matched EDA group in the model was 4308 Euros. It deviated from the base case result (5625 Euros) by 23% due to differences in effect but the costs were unchanged (Table 6). Discussion We have modelled the most relevant clinical pathways for a patient treated for postoperative pain with epidural or patient-controlled intravenous analgesia. Based on a clinical database we have described how these two alternatives actually work under clinical circumstances and not how they ideally can work. The approach to express the cost-effectiveness In the health economic literature and methodological guidelines the costs per quality adjusted life year is recommended for use in CEA even if the gain is not a whole year. As we have analysed a very short effect duration of acute pain treatment and found that the cost per gained life year is inadequate, the cost per gained pain-free day was used. 115

6 Bartha et al. An economic analysis in health care should describe a favourable improvement in patient s health or experience, and therefore the number of pain-free days is not optimal, e.g. number of pain-free days based on VAS do not capture the multidimensional experience of pain; it is an intermediate measure of outcome. As the influence of improved pain relief on the postoperative outcome is a mater of debate, other measures of outcome should also be considered. Using EDA, compared with PCIA, the length and cost of postoperative care was reported to be less, 615 but the length of hospital stay was found to be the same or longer. 14 Another limitation in our study is that only one representative value of the pain score was registered in the database; however the variability of the sampled pain scores was low. Besides these limitations the sensitivity analysis shows that the ICER was sensitive to some changes in critical input values in the model, but in all tested alternatives the PCIA was more cost-effective than EDA. The higher cost of postoperative epidural analgesia compared with PCIA is in line with others but the range of difference varies. In those studies other measures of outcome were used, and the results were not expressed in terms of cost-effectiveness making the comparison with our result difficult. Our result is based on real clinical data which allow some generalization, but use of more sophisticated methods for epidural analgesia or a separate analysis of low and high laparotomy may influence the result. However the probabilities of early discontinuation of treatment, accidental dislocation of catheter, minor side-effects or difficulties with catheter insertion are not likely to be affected by these. Why a model analysis? When alternative treatments are already established, ethical considerations may preclude randomized trials and consecutive patient selection may be preferred, as in a recent Swedish trial. 17 Furthermore costs and effects evaluated in an RCT may be difficult to transform into clinical practice, because of the influence of controlled experimental settings, and the possible lack of clinical endpoints (e.g. effect-data of discontinued treatment) that are relevant for economic evaluation. 18 The clinical endpoints for discontinued pathways of treatment with EDA were documented in our database, and therefore it was considered to be an important data source. Implication of our results and conclusion Bearing in mind the limitations of this study we conclude with reasonable degree of certainty that the cost of postoperative epidural analgesia is three times higher than that of PCIA but also that it gives better pain relief than PCIA, which shows that PCIA probably is the most costeffective alternative under the actual circumstances. Decreasing the cost of the postoperative care and/or diminishing the technical difficulties related to epidural analgesia may cut the costs for EDA. A model (decision-tree) makes the different clinical pathways transparent, and it allows evaluation of assumed improvements. Moreover the structured analysis of the particular clinical pathways may give additional information, especially when the results of RCTs are difficult to summarize, and there is still doubt as to which technique is better in terms of analgesia, adverse effects and benefits. 19 The better choice is not always the best one if resources are scarce, and it may be right to go for the second best alternative. According to Swedish law patients have the right to choose a more expensive treatment if the cost is reasonable in relation to the benefit. It is a judgement of value whether the additional cost of 5625 Euros per painfree day is reasonable for society to pay for the obtained patient benefit. It is a question of opportunity costs, i.e. who else will suffer because of lack of health-care resources? In a broader health policy context, when the cost of around Euros per extra life-year gained with full health (QALY) for other interventions are questioned, our results indicate a poor cost-effectiveness. We have to go for good control (for example tracing the clinical pathways with decision-tree) to be able to make improvements both in terms of costs and effects. Before any conclusion can be drawn concerning policy recommendations the difference in costs has to be related to other outcome measures such as length of hospital stay, morbidity, and mortality and also we need other outcome measures that reflect the improved experience or health state of patients due to better pain relief. Acknowledgements The authors are grateful to Mona Lindblad and Lilian Adamson, Department of Anaesthesiology, University Hospital Linköping, Sweden for help with data acquisition and Martin Henriksson CMT/HIS Linköping University for valuable comments on the manuscript. References 1 Brodner G, Van Aken H, Hertle L, et al. Multimodal perioperative management combining thoracic epidural analgesia, forced mobilization, and oral nutrition reduces hormonal and metabolic stress and improves convalescence after major urologic surgery. Anesth Analg 2001; 92: Kehlet H. Acute pain control and accelerated postoperative surgical recovery. Surg Clin North Am 1999; 79: Kehlet H, Mogensen T. Hospital stay of 2 days after open sigmoidectomy with a multimodal rehabilitation programme. Br J Surg 1999; 86: McLeod G, Davies H, Munnoch N, et al. Postoperative pain relief using thoracic epidural analgesia: outstanding success and disappointing failures. Anaesthesia 2001; 56: Andersen G, Rasmussen H, Rosenstock C, et al. Postoperative pain control by epidural analgesia after transabdominal surgery. Efficacy and problems encountered in daily routine. Acta Anaesthesiol Scand 2000; 44: Brodner G, Mertes N, Buerkle H, et al. Acute pain management: analysis, implications and consequences after prospective experience with 6349 surgical patients. Eur J Anaesthesiol 2000; 17: Wigfull J, Welchew E. Survey of 1057 patients receiving postoperative patient-controlled epidural analgesia. Anaesthesia 2001; 56:

7 Economic evaluation of postoperative epidural and patient-controlled intravenous analgesia 8 Drummond MF, O Brien BJ, Stoddart GL, Torrance GW. Methods for the Economical Evaluation of Health Care Programmes, 2nd edn. New York, Oxford: Oxford University Press, 1997; 117, 120, 242, 276, Karlsson G, Johannesson M. The decision rules of costeffectiveness analysis. Pharmaeconomics 1996; 9: Drummond MF, O Brien BJ, Stoddart GL, Torrance GW. Methods for the Economical Evaluation of Health Care Programmes, 2nd edn. New York, Oxford: Oxford University Press, 1997; Rockemann MG, Seeling W, Goertz AW, et al. Effectiveness, side effects and costs of postoperative pain therapy: intravenous and epidural patient-controlled analgesia (PCA). Anasthesiol Intensivmed Notfallmed Schmerzther 1997; 32: Paulsen EK, Porter MG, Helmer SD, et al. Thoracic epidural versus patient-controlled analgesia in elective bowel resections. Am J Surg 2001; 182: Cohen BE, Hartman MB, Wade JT, et al. Postoperative pain control after lumbar spine fusion. Patient-controlled analgesia versus continuous epidural analgesia. Spine 1997; 22: ; discussion Ammar AD. Postoperative epidural analgesia following abdominal aortic surgery: do the benefits justify the costs? Ann Vasc Surg 1998; 12: de Leon-Casasola OA, Parker BM, Lema MJ, et al. Epidural analgesia versus intravenous patient-controlled analgesia. Differences in the postoperative course of cancer patients. Reg Anesth 1994; 19: Norris EJ, Beattie C, Perler BA, et al. Double-masked randomized trial comparing alternate combinations of intraoperative anesthesia and postoperative analgesia in abdominal aortic surgery. Anesthesiology 2001; 95: Flisberg P, Rudin A, Linner R, et al. Pain relief and safety after major surgery. A prospective study of epidural and intravenous analgesia in 2696 patients. Acta Anaesthesiol Scand 2003; 47: Drummond MF, O Brien BJ, Stoddart GL, Torrance GW. Methods for the Economical Evaluation of Health Care Programmes, 2nd edn. New York, Oxford: Oxford University Press, 1997; Werawatganon T, Charuluxanun S. Patient controlled intravenous opioid analgesia versus continuous epidural analgesia for pain after intra-abdominal surgery. Cochrane Database Syst Rev 2005; 1: CD