Abstract
Background: Accelerated perioperative rehabilitation protocols following total hip and knee arthroplasties are currently being implemented worldwide, but the cost-effectiveness of these protocols from a societal perspective is not known. We compared the cost-effectiveness of an accelerated perioperative care and rehabilitation protocol with that of a more standard protocol for patients treated with total hip and knee arthroplasty.
Methods: A cost-effectiveness study was undertaken as a study piggybacked on a randomized clinical trial comparing early outcomes of an accelerated and intensive postoperative rehabilitation regimen with those of a more standard rehabilitation protocol. We assessed eighty-seven patients (forty-two who received the standard protocol and forty-five who received the accelerated protocol) for a total of twelve months. Costs from the time of the patient's visit immediately before the operation to one year postoperatively were calculated with use of activity-based costing analysis. Postoperative quality-adjusted life-years (QALYs) were calculated from validated patient diaries and questionnaires at fifteen time points. The primary objective was to determine whether one intervention was dominant over the other during a twelve-month period or, if neither was dominant, to determine the incremental cost-effectiveness ratio.
Results: The result of the randomized clinical trial showed the accelerated intervention to be effective, with a reduction in the length of the hospital stay and a gain in health-related quality of life at the three-month follow-up time point. The cost-effectiveness study showed the accelerated protocol to be significantly less expensive than the standard protocol (p = 0.036), with an average reduction in cost of 18,880 Danish kroner (95% confidence interval, 1899 to 38,152) (approximately US $4000). Patients treated with the accelerated protocol following hip arthroplasty had an additional average gain of 0.08 QALY (95% confidence interval, 0.02 to 0.15) compared with the patients who received the standard protocol (p = 0.006); this led to a 98% dominance of the accelerated protocol over the standard protocol. No significant or clinically relevant difference in the numbers of QALYs associated with the two protocols was observed for the patients treated with knee arthroplasty.
Conclusions: An accelerated perioperative care and rehabilitation protocol can be both cost-saving and clinically more effective after total hip arthroplasty, whereas it can be cost-saving with no observed significant difference in effect, from a societal perspective, after knee arthroplasty.
Level of Evidence: Economic and decision analysis Level I. See Instructions to Authors for a complete description of levels of evidence.
In Denmark, 12,000 total hip and knee arthroplasties are performed each year1. The total hospital costs for those procedures were approximately 700 million Danish kroner (DKK) (110 million U.S. dollars) in 20052. New protocols designed to optimize perioperative care and rehabilitation have been given several different names, such as "accelerated intervention" or "clinical pathway." An accelerated intervention is defined as a multimodal intervention taking place in a multidisciplinary or interdisciplinary organization in order to shorten the time to recovery3-7. In contrast, clinical pathways have been implemented in the United States in an effort to reduce the length of the hospital stay and thereby control hospital costs, with less focus placed on consequences for patients and society8.
While clinical pathways for hip and knee replacement have been shown to reduce the length of the hospital stay and costs9, their major impact is from a hospital perspective. However, to establish the cost-effectiveness of these interventions, they should be assessed from a societal perspective10. The efficacy of accelerated perioperative care and interventions has recently been established11, but to our knowledge there is no evidence regarding their cost-effectiveness. We therefore investigated, from a societal perspective, the cost-effectiveness of an accelerated perioperative care and rehabilitation intervention, compared with a more standard protocol, following total hip and knee arthroplasty.
A cost-utility study was piggybacked on a randomized clinical trial performed at the Orthopedic Clinic in the Regional Hospital Holstebro, Denmark, from August 2005 to February 200711. The randomized clinical trial was performed to study the effect of accelerated intervention, defined as specialized proactive care and early intensive mobilization and exercise, compared with that of a regimen that was less personalized and less intensive. The randomized clinical trial included eighty-seven patients treated with total hip arthroplasty or with unicompartmental or total knee arthroplasty. The outcome measures of the randomized clinical trial were the length of the stay in the hospital, gain in health-related quality of life from baseline to three months postoperatively, and adverse effects within the first three months postoperatively. The main results of that study were a reduction in the mean length of the hospital stay from eight days (95% confidence interval, 7.1 to 8.4) in the standard-protocol group to five days (95% confidence interval, 4.2 to 5.6) in the accelerated-protocol group (p < 0.001). This was accompanied by a larger gain in the health-related quality of life of 0.08 (95% confidence interval, 0.004 to 0.16) in the accelerated-protocol group (p = 0.03). No significant or clinically relevant differences in adverse effects were observed.
In the current cost-effectiveness study, we followed all eighty-seven patients for an additional nine months and focused on both costs and cost-effectiveness in the first year postoperatively. The procedures followed in this study were in accordance with the ethical standards of the responsible committee on human experimentation and with the Helsinki Declaration of 1975, as revised in 2000. The study protocol was approved by the Medical Ethics Committee of Ringkoebing and Southern Jutland Counties, and The Danish Data Protection Agency.
Patients
All patients who were scheduled to undergo elective primary total hip arthroplasty, total knee arthroplasty, or unicompartmental knee arthroplasty at our center were consecutively invited to participate in the study. All patients meeting the inclusion criteria were provided written and oral information about the study at the initial visit, and interested patients gave written consent. The inclusion criteria were a planned elective primary total hip arthroplasty, total knee arthroplasty, or unicompartmental knee arthroplasty. Exclusion criteria were neurological impairment to a degree that the patient could not understand the instructions for participation and a lack of written consent prior to hospitalization.
We determined that a cost saving resulting from a two-day reduction in the length of the hospital stay would be clinically relevant. We performed a power analysis and determined the need for at least forty patients in each group with alpha set at 0.05 and beta set at 0.80. Altogether, 117 patients were eligible for the study (Fig. 1). Twenty-three patients refused to participate, and those twenty-three patients had an average age (and standard deviation) of 67 ± 10.4 years, a female-to-male ratio of 14:9, and a total hip arthroplasty-to-total knee arthroplasty ratio of 10:13. An additional four patients did not meet the inclusion criteria. This left ninety patients for randomization. Forty-five patients were allocated to each group. Three patients in the standard-protocol group (two scheduled to be treated with total hip arthroplasty and one scheduled to be treated with knee arthroplasty) were excluded after randomization. One patient was excluded because the surgery was cancelled as a result of an infection, and two were excluded because they wanted the surgery to be performed outside of the inclusion period. This left eighty-seven patients to receive the allocated intervention. The patients were randomized to either a standard-intervention group or a new accelerated-intervention group by a secretary who was not otherwise involved in the study and who drew an opaque envelope from a box. After randomization, the patients filled in a baseline questionnaire to establish their preoperative health-related quality of life. Other data on patient characteristics were drawn from hospital registers. The patient characteristics of both groups are presented as an electronic appendix.
Protocol: Both Groups
The patients in both groups were treated with identical surgical and anesthetic procedures according to Danish guidelines, one of which is the use of cemented implants for total hip arthroplasties performed in patients over sixty-five years of age12,13. Five experienced surgeons performed all of the operations. Three surgeons performed total hip and total knee arthroplasties at a ratio of 2:1, the fourth surgeon performed only total hip arthroplasties, and the fifth surgeon performed total knee and total hip arthroplasties at a ratio of 9:1. The surgeons were equally represented in both intervention groups. Medications for pain relief were identical in the two groups and consisted of OxyContin or OxyNorm (oxycodone) and paracetamol; Zofran (ondansetron) was used for nausea reduction, and magnesia tablets were given for bowel regulation.
The standard and accelerated protocols are summarized in Table I. A detailed description of the accelerated protocol has been published previously14.
Discharge
All patients were discharged to home. The discharge criteria were the same for both groups: absence of any signs of wound problems; the patient's willingness to be discharged; satisfactory pain control; awareness of procedures for safely discontinuing the use of medication; knowledge of restrictions; the ability to walk safely with or without walking aids, to walk up and down stairs, and to perform home exercises; knowing how to increase home exercises; the ability to perform personal care; and, in the group treated with knee arthroplasty, achievement of at least 90° of knee flexion. Surgeons who were not otherwise involved in the study determined, in agreement with the patients, when the discharge criteria were fulfilled; no patient asked to postpone discharge after having fulfilled the other discharge criteria.
Attempts to Reduce Bias in the Study
The standard-protocol patient group and health-care staff were separated from the accelerated-protocol patient group and health-care staff during the study period, and the health-care staffs were not allowed to discuss the interventions with each other. Two newly employed therapists, a physiotherapist and an occupational therapist, blinded to the standard protocol were mainly responsible for the rehabilitation in the accelerated-protocol group. Health-care personnel carrying out the standard protocol were not aware of the procedures in the accelerated-protocol group. Because the length of the hospital stay was related to both the intervention and the outcome, surgeons who were not otherwise involved in the study decided when the discharge criteria were fulfilled. We used validated questionnaires15, diaries16, and registers17 for all data collection.
Economic Evaluation
The analysis was based on a societal perspective, and the time frame was fixed to one year per patient. The analysis was a marginal analysis (investigating only areas that were different). The cost-effectiveness of the accelerated intervention was estimated by relating the incremental cost of the two interventions to the incremental effect, in quality-adjusted life-years (QALYs), of the two interventions. The resulting incremental cost-effectiveness ratio represents the cost per QALY gained in a cost-utility analysis10. In cases where the new intervention was both less costly and more effective, an estimate of the percent dominance was obtained from a bootstrap simulation. The uncertainty of the incremental cost-effectiveness ratio was also estimated by using bootstrap simulation. A bootstrap simulation is a nonparametric method in which a random sample of the same size as the original sample is drawn several times with replacement from the original data. The results of the bootstrap sampling are presented in a cost-effectiveness plane. The cost-effectiveness plane is constructed from the crossing of the x axis and the y axis. Incremental effect is plotted on the x axis and incremental cost, on the y axis. The four resulting quadrants represent the potential outcomes in terms of cost and effect. In the upper right quadrant, the new intervention is more effective but also more costly than the comparator; in the lower right quadrant, it is more effective and less costly; in the lower left quadrant, it is less effective and less costly; and in the upper left quadrant, it is less effective and more costly10. A cost of DKK 160,000 per QALY gained18, a cost saving of at least DKK 1000 with no significant or clinically relevant difference in health-related quality of life, or a gain in health-related quality of life of at least 0.05 with no significant or clinically relevant difference in cost was considered sufficient for a decision of implementation.
Costs
The primary cost estimate was the average total cost (cost estimate A). Postoperative productivity loss was calculated with a maximum of three months of absence from paid work (friction cost method)19. We obtained all costs for medical care, medication, and physiotherapy during the twelve weeks prior to hospitalization from a regional register17 in order to adjust the postoperative costs for preoperative costs. We started estimation of costs on the information day (the day of the patient's visit to receive information about the protocol and for consultation with the surgeon, anesthetist, and nurse), four days before the surgery, for the accelerated-intervention group and on the day of admission, which was the day before the surgery, for the standard-intervention group. Time ended at the one-year follow-up visit, at which a questionnaire was used. We estimated the total costs at the patient level using a mix of activity-based costing analysis (time studies) and the step-down method of allocation of overhead costs to all departments involved in the patient's care up to the final department (the orthopaedic ward). To cover the perioperative and postoperative patient path, we defined seven cost activity centers: (1) information day, (2) boarding of the patient at the hospital, (3) care in the hospital, (4) rehabilitation in the hospital, (5) patient needs (e.g., nonprescription medication, home changes, and transportation) in the follow-up period, (6) primary care in the follow-up period, and (7) hospital readmission in the follow-up period. The total average costs in the seven activity centers were calculated by multiplying the observed volumes of health care by the unit prices. Information-day activities were identified by means of observation and time registration and were validated by health-care staff. The costs of boarding the patient at the hospital were calculated on the basis of data from the hospital central accounting system with use of a step-down allocation of overhead costs to all departments involved in the patient's care up to the final department and by finally estimating a daily hospital-board cost. Care and rehabilitation activities in the hospital were identified by means of observation and time registration and were validated by health-care staff. Patient needs and primary care in the follow-up period were identified from standardized and validated patient diaries16 and questionnaires15 and were validated against the hospital patient administrative system and from a register of primary care data in a closed database on the Internet17. The register of primary care data holds data on all medical care consultations, prescribed medication, and physiotherapy consultations18. The patient diary was handed out on the day of discharge and consisted of a cost questionnaire, to be filled in weekly, regarding all relevant post-hospitalization costs over twelve weeks. The patients brought the diary to the three-month inpatient visit. If they failed to bring the diary or it was incomplete, the patients filled in a retrospective questionnaire or data were obtained by use of the register of primary care data17. Cost questionnaires were sent to the patients on the twenty-sixth, thirty-ninth, and fifty-second weeks. If there was a delay in the return of a questionnaire, a new questionnaire was sent, and if that questionnaire was not returned, data were obtained from the register of primary care data17. All postoperative patient and primary-care costs were therefore collected for a total of twelve months. Finally, hospital readmissions in the period from discharge to three months after discharge were identified on the basis of data from the hospital patient administrative system.
Preoperative and perioperative cost data were available for all patients. Eight patients (three treated with total hip arthroplasty and two treated with total knee arthroplasty followed by the standard intervention, and two treated with total hip arthroplasty and one treated with total knee arthroplasty followed by the accelerated intervention) did not submit a postoperative diary, and three patients (two treated with total hip arthroplasty followed by the standard intervention and one treated with total hip arthroplasty followed by the accelerated intervention) did not complete any of the postoperative questionnaires. However, data for these patients were obtained with retrospective questionnaires and from the register of primary care data17.
Unit Costs
Unit costs were obtained from the central Danish hospital employee register20, the register of primary care data17, StatBank Denmark21, the Dutch Manual for Costing in Economic Evaluations 200222, and patient reporting. The average number of effective working hours was calculated to be 1516 hours by using actual hospital wage and employee data. Costs in activity centers 1 through 4 were calculated in 2005 prices and transformed to 2006 prices after adjusting for inflation. Costs in activity centers 5, 6, and 7 were calculated with use of 2006 prices. Productivity loss for patients engaged in active employment was calculated by using an average wage rate for the age-specific group, while productivity loss for patients not engaged in active employment was calculated with use of the proposed tariff in the Dutch Manual for Costing in Economic Evaluations22, after adjusting for inflation.
Effects
The effect of the accelerated intervention was determined by using a standardized questionnaire for the measurement of health outcome, the EuroQOL-5D (EQ-5D)15. The EQ-5D measures health-related quality of life in five dimensions (mobility, self-care, usual activities, pain/discomfort, and anxiety/depression)15. The EQ-5D questionnaire was included in the patient diary and was filled in at baseline and then weekly from the first to twelfth week postoperatively. It was also completed at twenty-six, thirty-nine, and fifty-two weeks postoperatively. If return of a questionnaire was delayed, a new questionnaire was sent, and if that questionnaire was not returned, we contacted the patients by telephone at fifty-two weeks. Health-related quality-of-life scores were calculated with use of the "Official Danish Time Trade-Off Scores."23 The effect in terms of QALYs was calculated by using the fifteen measurement points postoperatively and employing the area-under-the-curve method with the trapezoid rule24. These effect data were available for all patients at baseline and at three and twelve months postoperatively, with the exception of the twelve-month follow-up data for one patient, who could not be reached by mail or telephone.
Statistical Methods
The primary analysis was a univariate analysis of incremental cost and effects. It was used because of the expectation of some degree of non-normality of the data, estimated by employing a nonparametric bootstrap procedure with 2000 bias-corrected bootstrap replicates of the arithmetic mean. Missing values resulting from incompleteness of data were, in accordance with the method described by Brunenberg et al.25, replaced with the mean value for the group in the univariate analysis of health-related quality of life and with regression imputation in the multivariate analysis. A total of 1.2% of the data points were imputed. The significance level was set at p < 0.05 and tested with the nonparametric percentile method24.
Uncertainty of Results Due to Choice of Model and Cost Estimate
Different Analysis Models
We performed multivariate analyses of incremental costs because this method can be superior to univariate analysis as a result of its ability to explain variation due to other causes24. Primary multivariate analysis of incremental costs (cost estimate A) and effects was performed by using ordinary least-square regression with 2000 bias-corrected and accelerated replicates of the incremental difference. We also performed secondary multivariate analyses of incremental costs with generalized linear models because these models seem to be ideal for handling the mean and variance functions on the original scale of skewed cost data24,26. Generalized linear model analyses were performed with a log link function and the following families: Gaussian, Poisson, Gamma, and inverse Gaussian/Wald. We report only the generalized linear model link functions and families that passed all of the following tests: skewness/kurtosis test, heteroskedasticity test (Breusch-Pagan test), modified Park test (generalized linear models family test), Pregibon link test, modified Hosmer-Lemeshow test, and Pearson correlation test24. Differences between treatment groups were tested by using the nonparametric percentile method24. In the multivariate analyses, incremental cost was adjusted for any preoperative primary-care cost, health-related quality of life at baseline, sex, age, a diagnosis of osteoarthritis or not, a cemented implant or not, whether the patient was employed or not, and whether a hip or knee arthroplasty was performed. In the multivariate analyses of incremental effect, we adjusted for health-related quality of life at baseline together with sex, age, a diagnosis of osteoarthritis or not, a cemented implant or not, and whether a hip or knee arthroplasty was performed.
Different Cost Estimates
Analyses with an additional three cost estimates (cost estimates B, C, and D) were carried out to enhance the transferability of costs in different areas within or outside the hospital and to evaluate their consequences for the conclusion. Cost estimate B was the total average cost with exclusion of the average costs of hospital readmissions in the follow-up period and was estimated because our results could be heavily affected by some fortuitous readmissions in one of the randomization arms. Cost estimate C was the total average cost with exclusion of the average productivity loss and was estimated because productivity loss is considered to be a separate cost group22. Cost estimate D was the total average cost with exclusion of both the average hospital-readmission costs and the average productivity loss.
Source of Funding
There was no external funding for this study.
Patients
There were forty-two patients (twenty-eight treated with total hip arthroplasty; twelve, with total knee arthroplasty; and two, with unicompartmental knee arthroplasty) in the standard-protocol group and forty-five patients (twenty-eight treated with total hip arthroplasty; fifteen, with total knee arthroplasty; and two, with unicompartmental knee arthroplasty) in the accelerated-protocol group. There were no significant differences between the groups regarding patient characteristics at baseline or the proportions of total hip arthroplasties, total knee arthroplasties, and unicompartmental knee arthroplasties (see Appendix). In addition, there were no significant differences in the baseline health-related quality of life or the postoperative number of QALYs between the intervention groups in the total series (patients undergoing total hip arthroplasty, total knee arthroplasty, and unicompartmental knee arthroplasty). We did, however, observe a marginally significant difference (defined as p < 0.1) between the two intervention groups with regard to the baseline health-related quality of life of patients undergoing total knee arthroplasty or unicompartmental knee arthroplasty (p = 0.053) and with regard to whether the implant had been cemented in the series as a whole (p = 0.082) (see Appendix). The Spearman rho correlation between the baseline health-related quality of life and the postoperative number of QALYs in the group treated with total knee arthroplasty or unicompartmental knee arthroplasty was estimated to be 0.21 (p = 0.249).
One patient treated with total hip arthroplasty who had been assigned to the standard-protocol group died on the day after the surgery because of a pulmonary embolism. Three patients were readmitted to the hospital within three months after discharge. In the standard-protocol group, one patient treated with total knee arthroplasty was readmitted because of a wound infection. This patient underwent revision surgery, which increased the total length of the hospital stay by fifteen days. In the accelerated-protocol group, two patients were readmitted to the hospital: one patient who had been treated with total knee arthroplasty spent an additional eleven days in the hospital because of swelling and pain in the knee, and one patient who had been treated with total hip arthroplasty spent an additional day in the hospital because of dislocation of the hip. This led to the addition of twelve days to the total length of the hospital stay in the accelerated-protocol group. Thirteen patients (six treated with total hip arthroplasty and one treated with total knee arthroplasty in the accelerated-protocol group and five treated with total hip arthroplasty and one treated with total knee arthroplasty in the standard-protocol group) received a second arthroplasty, in the contralateral hip or knee, during the study period after they had recovered from the first operation. In the accelerated-protocol group, four patients underwent the second total hip arthroplasty three to six months after the index procedure, whereas only one patient in the standard-protocol group had a second arthroplasty in this time period.
Costs
The average total cost (and standard deviation) was DKK 90,227 ± 47,475 (approximately US $18,000) in the standard-protocol group and DKK 71,344 ± 39,958 (approximately US $14,000) in the accelerated-protocol group. The average costs, grouped according to the seven cost activity centers, for the patients who underwent total hip arthroplasty and for those who underwent total knee arthroplasty or unicompartmental knee arthroplasty are presented in a table in the Appendix. The average total cost for the patients treated with total hip arthroplasty was DKK 87,657 ± 39,915 in the standard-protocol group and DKK 71,768 ± 41,544 in the accelerated-protocol group. The average total cost for the patients treated with total knee arthroplasty or unicompartmental knee arthroplasty was DKK 95,367 ± 61,293 in the standard-protocol group and DKK 70,644 ± 38,437 in the accelerated-protocol group. For the patient treated with total knee arthroplasty in the standard-protocol group who was readmitted for a reoperation, the hospital-readmission cost was DKK 7983 (approximately US $1600). The univariate analysis showed an incremental average total cost of DKK -18,880 (95% confidence interval, -1899 to -38,152) (approximately US $-4000) in favor of the accelerated intervention (p = 0.036) (Table II). No significant interaction between the intervention and randomization stratification groups was observed for costs.
Effects
The health-related quality of life increased on a weekly basis from the first week postoperatively and peaked at different times in the four treatment groups (Figs. 2 and 3). The health-related quality of life peaked at twelve weeks for patients treated with total hip arthroplasty followed by the accelerated intervention, at thirty-nine weeks for those treated with total knee arthroplasty or unicompartmental knee arthroplasty irrespective of the postoperative intervention, and at fifty-two weeks for patients treated with total hip arthroplasty followed by the standard intervention. The average number of QALYs was 0.78 ± 0.15 in the standard-protocol group compared with 0.83 ± 0.10 in the accelerated-protocol group. The univariate crude analysis revealed a significant and clinically relevant incremental effect, in QALYs, of 0.05 (95% confidence interval, 0.01 to 0.12), favoring the accelerated intervention (p = 0.029) (see Appendix). In the group with total hip arthroplasty, the mean number of QALYs was 0.75 ± 0.18 for those treated with the standard protocol compared with 0.84 ± 0.11 for those treated with the accelerated protocol. In the group with knee arthroplasty, the mean number of QALYs was 0.85 ± 0.05 for those treated with the standard protocol and 0.81 ± 0.09 for those treated with the accelerated protocol. However, a significant interaction between the intervention and randomization stratification groups was observed for effect (p = 0.042). The multivariate stratified analysis of incremental effect revealed that the results observed in the crude unadjusted analysis were actually due to an effect contribution from patients treated with total hip arthroplasty, with a significant and clinically relevant incremental effect of 0.08 (95% confidence interval, 0.02 to 0.15) favoring accelerated intervention (p = 0.006). With the numbers studied, no significant or clinically relevant difference in effect could be observed for the patients treated with knee arthroplasty (see Appendix).
Cost-Effectiveness
The results of our univariate crude analysis of 2000 bootstrap replicates showed that the accelerated intervention was both significantly less costly and significantly more effective than the standard intervention. A total of 97% of the cost-effectiveness pairs were placed in the lower right corner of the cost-effectiveness plane (Fig. 4). Because of a significant difference in the effect of the intervention between the patients treated with total hip arthroplasty and those treated with total knee arthroplasty or unicompartmental knee arthroplasty, we also present the results of multivariate stratified cost-effectiveness analyses separately for the patients treated with total hip arthroplasty and those treated with knee arthroplasty. For the patients treated with total hip arthroplasty, the accelerated intervention dominated the standard intervention (98% of the cost-effectiveness pairs were placed in the lower right corner of the cost-effectiveness plane) (Fig. 5). For the patients treated with knee arthroplasty, we observed less cost in association with the accelerated intervention (93% of the cost-effectiveness pairs were placed in the lower half of the cost-effectiveness plane) but no significant or clinically relevant difference in effect (Fig.6).
Uncertainty of Results Due to Choice of Model and Cost Estimate
Irrespective of the analysis model or type of cost estimate, the results consistently favored the accelerated intervention in terms of costs. In the stratified analyses of costs, the incremental cost estimates calculated with the ordinary least-square regression differed by approximately DKK 2900 (approximately US $600) from the estimates derived with the generalized linear model analyses. With regard to the different cost estimates, exclusion of hospital-readmission costs led to almost no change in the incremental cost for the patients treated with total hip arthroplasty but a large decrease for the patients treated with total knee arthroplasty or unicompartmental knee arthroplasty. Exclusion of costs from productivity loss led to a large decrease in incremental cost for the patients treated with total hip arthroplasty, whereas the result was almost constant for the patients treated with total knee arthroplasty or unicompartmental knee arthroplasty (see Appendix).
To our knowledge, this is the first cost-effectiveness study based on a randomized clinical trial comparing an accelerated perioperative care and rehabilitation intervention with a standard regimen after hip and knee arthroplasty. We found that, from a societal perspective, accelerated intervention dominates the standard intervention when used for patients treated with hip arthroplasty and is less costly without adverse effects for those treated with knee arthroplasty.
The costs in all cost activity centers, except for follow-up primary care and hospital readmission for patients treated with total hip arthroplasty and information day for both groups, were consistently lower in the accelerated-intervention group than in the standard-intervention group. This could indicate some degree of cost shifting from the hospital to the primary-care sector. The results of our univariate analysis showed the costs in the accelerated-intervention group to be almost DKK 19,000 (approximately US $4000) less than those in the standard-intervention group, with an additional gain of around 0.05 QALY in the accelerated-intervention group. The most common finding in cost-effectiveness studies is that the new intervention is better but also more expensive than the comparator. One therefore has to develop certain decision rules to be able to decide if one would adopt the new intervention or not. The most frequently used decision rules are that a new intervention should be adopted if the cost is less than US $50,000 per QALY gained in the United States24, less than £20,000 per QALY gained in the United Kingdom27, and less than DKK 160,000 per QALY gained in Denmark18. However, in our study, no decision rule or threshold of cost per QALY gained came into play because accelerated intervention was found to be the dominant strategy for patients treated with total hip arthroplasty and to result in a cost saving with no significant or clinically relevant difference in effect for patients treated with total knee arthroplasty or unicompartmental knee arthroplasty. The cost saving for the patients treated with total knee arthroplasty or unicompartmental knee arthroplasty was much greater than our threshold of DKK 1000.
In accordance with the recommendations by Glick et al.24, we also performed a multivariate analysis together with different analysis models and different cost estimates to provide the reader with a degree of variation in the estimates due to model and estimate uncertainty. The results from these analyses did not change the overall conclusion, but the estimates did vary. The multivariate analysis identified a significant interaction between the randomization stratification groups and the intervention groups, indicating that a stratified analysis would be the most appropriate. The method of cost estimation also affected the results to some degree. When we excluded hospital-readmission costs from the cost estimates, it did not affect the results for the patients treated with total hip arthroplasty but it did affect the results for the patients treated with total knee arthroplasty or unicompartmental knee arthroplasty because of one patient in the standard-protocol group who underwent a second operation. About half of the incremental cost for the patients treated with total hip arthroplasty was explained by a difference in productivity loss between the two intervention groups. The patients treated with total knee arthroplasty or unicompartmental knee arthroplasty had a different pattern regarding productivity loss, mostly because of a lower general proportion of people who were employed but also because of a difference in the proportions of employed people between the standard and accelerated-intervention groups. In the stratum-specific analysis of effect, the results favoring the accelerated intervention were solely due to an effect gain by the patients treated with total hip arthroplasty. Our results therefore demonstrate the usefulness of different analysis models and estimates of transferability.
Regarding the costs for the patients treated with the accelerated intervention, our findings are in accordance with those of the only other published study on this topic of which we are aware25. In that study, the authors used a similar intervention, but in a non-randomized-design trial. The incremental costs in both studies favored the new, accelerated intervention, although the difference was more pronounced in our study. The postoperative costs in both groups in our study were higher than those in the other study25; this difference can be primarily explained by a greater proportion of employed people in our study.
The health-related quality of life that was observed in our study was much higher, in all patient groups and at all comparable time points, than that reported by Brunenberg et al.25. Some of this difference could be explained by differences in the indications for surgery between the two countries (The Netherlands and Denmark). Both in the study by Brunenberg et al. and in our study, the patients treated with total hip arthroplasty followed by the accelerated protocol were observed to have a decline in their health-related quality of life between six and twelve months postoperatively. In our study, some of this decline could be ascribed to patients in the accelerated-protocol group who regained their health more quickly and therefore received a second total hip arthroplasty sooner, leading to a second recovery period. In a multivariate analysis, Brunenberg et al. demonstrated a clinically relevant but non-significant incremental effect of 0.05 QALYs favoring the accelerated intervention in the group treated with hip arthroplasty. We demonstrated an incremental effect of the same magnitude, and that was also significant, for patients treated with hip arthroplasty. Neither study identified a significant difference in the number of QALYs between the two groups of patients treated with knee arthroplasty. However, it should be noted that our point estimate for the group treated with knee arthroplasty actually was placed in the lower left quadrant in the cost-effectiveness plane (Fig. 6) and therefore could indicate both less cost but also less effectiveness as compared with the standard rehabilitation protocol. This potential problem has to be addressed in future research in the area.
Brunenberg et al.25 demonstrated that accelerated intervention was superior to the standard intervention after both hip and knee arthroplasties. We were able to demonstrate superiority only for the patients treated with hip arthroplasty. Altogether, we observed adverse effects (one death and three readmissions) in four patients. This finding is in accordance with reports from other Danish hospitals1. However, our study was not designed to examine differences in adverse effects. Further monitoring of adverse effects has to be performed if accelerated procedures are implemented.
We took great effort to minimize bias during the study period11. We reduced selection bias through consecutive inclusion of patients and by using broad inclusion criteria, and we excluded patients only if they were so neurologically impaired that they could not understand the instructions for participation. There was, however, some potential imbalance, after randomization, between the two intervention groups with regard to the baseline health-related quality of life of the patients treated with total knee arthroplasty or unicompartmental knee arthroplasty. This potential imbalance could be one explanation for the observed difference in health-related quality of life between the two intervention groups from the twenty-sixth week onward, and we did actually observe a low positive, but non-significant, correlation between baseline health-related quality of life and the postoperative number of QALYs in our sample. This observation is in agreement with the findings in the study by Ethgen et al., who showed that a high preoperative health-related quality of life predicts a high postoperative health-related quality of life in this patient group28. The observed potential imbalance between the intervention groups in terms of implant fixation (with or without cement) was of less concern because the difference in the knee arthroplasty group had no clinical relevance (and was recorded only because of coding procedure), whereas the types of implant fixation used for the total hip arthroplasties, which do matter clinically, were comparable at baseline. We believe that our attempts to mask the patients and the health-care personnel were also successful and thereby minimized bias. We also believe that we successfully minimized a potential observer bias by using questionnaires, diaries, and register data. Furthermore, we think that the discharge procedures were identical in both groups. We do not believe that using the criterion of patient acceptance of discharge created any bias between the two intervention groups because all patients accepted discharge on fulfillment of all other discharge criteria. We included only readmissions that took place within three months postoperatively because that is the period for registering readmissions as proposed by the administrators of the Danish hip and knee arthroplasty registers29. However, most readmissions occur within the first postoperative month1. In the multivariate analyses of incremental effect, we adjusted, in accordance with the method used by Manca et al.30, for health-related quality of life at baseline together with all other potential confounders or risk factors. We believe that the chosen time frame of one year includes the relevant costs and effects of the standard and accelerated-rehabilitation protocols. However, with 23% of the eligible patients not participating in the study, it is still uncertain whether cost-effectiveness at a population level during routine circumstances, especially with respect to productivity loss, can be demonstrated.
One of the strengths of our study is that it was performed by using a pragmatic design in a local hospital, and not with highly selected patients and health-care staff in a university hospital. Inclusion criteria were broad, and <5% of the patients were excluded. We therefore believe that if our intervention is implemented, similar results could be achieved in other hospitals and for society as a whole.
On the basis of our findings and those of the study by Brunenberg et al.25, we now believe that the available evidence favors the use of accelerated interventions after primary total hip arthroplasty, total knee arthroplasty, and unicompartmental knee arthroplasty.
Tables showing patient characteristics at baseline, results of activity-based costing analysis, results of univariate and multivariate analysis of effect, and results of different multivariate analysis models of costs are available with the electronic versions of this article, on our web site at (go to the article citation and click on "Supplementary Material") and on our quarterly CD/DVD (call our subscription department, at 781-449-9780, to order the CD or DVD). 
Note: The authors thank Hans Schmidt, Regional Hospital Holstebro, for his help in providing cost data for the activity-based cost analysis. They also thank the statisticians, Morten Frydenberg and Niels Trolle Andersen, University of Aarhus, Denmark, for their help in producing do-files for statistical analysis. Finally, they thank Henry A. Glick, University of Pennsylvania, for his available do-files for cost-effectiveness analysis.
Danish National Board of Health, Center for Evaluation and Health Technology Assessment. Evaluation of patient path for patients undergoing primary total hip or knee arthroplasty. Danish National Board of Health, Danish Centre for Evaluation and Health Technology Assessment, Copenhagen. 2006. . Accessed 2009 Feb 18.
2006
http://www.sst.dk/publ/Publ2006/CEMTV/Alloplastik/Hovedrapport.pdf
Kehlet H, Dahl JB. Anaesthesia, surgery, and challenges in postoperative recovery. Lancet.2003;362:1921-8.3621921
2003
[PubMed][CrossRef]
Kehlet H, Wilmore DW. Multimodal strategies to improve surgical outcome. Am J Surg.2002;183:630-41.183630
2002
[CrossRef]
Kehlet H, Wilmore DW. Fast-track surgery. Br J Surg.2005;92:3-4.923
2005
[CrossRef]
Wilmore DW, Kehlet H. Management of patients in fast track surgery. BMJ.2001;322:473-6.322473
2001
[CrossRef]
Thylefors I, Persson O, Hellström D. Team types, perceived efficiency and team climate in Swedish cross-professional teamwork. J Interprof Care.2005;19:102-14.19102
2005
[CrossRef]
Healy WL, Iorio R, Ko J, Appleby D, Lemos DW. Impact of cost reduction programs on short-term patient outcome and hospital cost of total knee arthroplasty. J Bone Joint Surg Am.2002;84:348-53.84348
2002
Kim S, Losina E, Solomon DH, Wright J, Katz JN. Effectiveness of clinical pathways for total knee and total hip arthroplasty: literature review. J Arthroplasty.2003;18:69-74.1869
2003
[CrossRef]
Drummond MF, O'Brien B, Stoddart GL, Torrance GW. Methods for the economic evaluation of health care programmes. 2nd ed. Oxford: Oxford University Press; 1997.
1997
Larsen K, Sørensen OG, Hansen TB, Thomsen PB, Søballe K. Accelerated perioperative care and rehabilitation intervention for hip and knee replacement is effective: a randomized clinical trial involving 87 patients with 3 months of follow-up. Acta Orthop.2008;79:149-59.79149
2008
[CrossRef]
Szende A, Williams A, editors. Measuring self-reported population health: an international perspective based on EQ-5D. Budapest: SpringMed; 2004.
2004
Goossens ME, Rutten-van Mölken MP, Vlaeyen JW, van der Linden SM. The cost diary: a method to measure direct and indirect costs in cost-effectiveness research. J Clin Epidemiol.2000;53:688-95.53688
2000
[CrossRef]
Ringkøbing Amts Sygesikringsregister. Ringkøbing Amt. [Local primary health care register not available on the internet.] Data accessed 2007 Jun 4.
2007
Koopmanschap MA, Rutten FF, van Ineveld BM, van Roijen L. The friction cost method for measuring indirect cost of disease. J Health Econ.1995;14:171-89.14171
1995
[CrossRef]
Oostenbrink JB, Koopmanschap MA, Rutten FF. Standardisation of costs: the Dutch Manual for Costing in Economic Evaluations. Pharmacoeconomics.2002;20:443-54.20443
2002
[CrossRef]
Pedersen KM, Wittrup-Jensen K, Brooks R, Gudex C. Værdisætning af sundhed. Odense, Denmark: Syddansk Universitetsforlag; 2003.
2003
Glick HA, Doshi JA, Sonnad SS, Polsky D. Economic evaluation in clinical trials. Oxford: Oxford University Press; 2007.
2007
Brunenberg DE, van Steyn MJ, Sluimer JC, Bekebrede LL, Bulstra SK, Joore MA. Joint recovery programme versus usual care: an economic evaluation of a clinical pathway for joint replacement surgery. Med Care.2005;43:1018-26.431018
2005
[CrossRef]
Willan AR, Briggs AH. Statistical analysis of cost-effectiveness data. Chichester, England: John Wiley; 2006.
2006
National Institute of Clinical Excellence. Guide to the methods of technology appraisal—no. NO515. London: National Institute of Clinical Excellence; 2004.
2004
Ethgen O, Bruyère O, Richy F, Dardennes C, Reginster JY. Health-related quality of life in total hip and total knee arthroplasty. A qualitative and systematic review of the literature. J. Bone Joint Surg Am.2004;86:963-74.86963
2004
Manca A, Hawkins N, Sculpher MJ. Estimating mean QALYs in trial-based cost-effectiveness analysis: the importance of controlling for baseline utility. Health Econ.2005;14:487-96.14487
2005
[CrossRef]