If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Address correspondence and reprint requests to: S.S. Tan, Erasmus University Rotterdam, Institute for Medical Technology Assessment & Institute of Health Policy and Management, P.O. Box 1738, 3000 DR Rotterdam, The Netherlands.
Erasmus University Rotterdam, Institute for Medical Technology Assessment & Institute of Health Policy and Management, P.O. Box 1738, 3000 DR Rotterdam, The NetherlandsErasmus MC University Medical Center, Department of Rehabilitation Medicine, Rotterdam, The Netherlands
To determine the cost-effectiveness (CE) of exercise therapy (intervention group) compared to ‘general practitioner (GP) care’ (control group) in patients with hip osteoarthritis (OA) in primary care.
This cost-utility analysis was conducted with 120 GPs in the Netherlands from the societal and healthcare perspective. Data on direct medical costs, productivity costs and quality of life (QoL) was collected using standardised questionnaires which were sent to the patients at baseline and at 6, 13, 26, 39 and 52 weeks follow-up. All costs were based on Euro 2011 cost data.
A total of 203 patients were included. The annual direct medical costs per patient were significantly lower for the intervention group (€ 1233) compared to the control group (€ 1331). The average annual societal costs per patient were lower in the intervention group (€ 2634 vs € 3241). Productivity costs were higher than direct medical costs. There was a very small adjusted difference in QoL of 0.006 in favour of the control group (95% CI: −0.04 to +0.02).
Our study revealed that exercise therapy is probably cost saving, without the risk of noteworthy negative health effects.
. The general practitioner (GP) is the initial caregiver involved, in many cases providing the patient with education and counselling, pain medication, referral to an orthopaedic surgeon and/or additional diagnostic examinations (GP care). As exercise therapy has been shown to reduce pain and improve physical functioning in patients with hip OA
Although recommendations for the treatment of hip OA are mainly based on knee OA studies, few studies have previously evaluated the effectiveness of exercise therapy in hip OA. Hernandez-Molina et al. reported that the most effective therapeutic exercise involves regular aerobic activity and/or a strengthening program
. As the beneficial effects of exercise therapy lasting up to 12 weeks seem to decline and eventually disappear, adding booster exercise sessions to the intervention is suggested to induce long-term effects
. Economic evaluations are a prerequisite for the reimbursement and implementation of treatments in many countries, because they can provide healthcare decision makers with valuable information on the relative efficiency of alternative treatments
Four earlier studies have assessed the cost-effectiveness (CE) of exercise therapy in patients with hip or knee OA. Reported costs per quality-adjusted life year (QALY) varied between about −$500 and $20.000
. Because of the lack of information on the costs as well as on the effectiveness of exercise therapy added to GP care (compared to GP care alone), GPs lack the knowledge to apply the most cost effective treatment to patients with hip OA. Therefore, the aim of the present study was to determine the CE of exercise therapy added to GP care (intervention group) compared to GP care only (control group) in patients with hip OA.
This cost-utility study was performed in conjunction with a randomised controlled trial. More details of the study design can be read in the protocol
. In short, patients were identified via patient registries of the participating GPs. Patients of ≥45 years with hip OA who have consulted their GP during the past year regarding a new episode of non-traumatic hip complaints and who complied with the clinical American College of Rheumatology criteria for hip OA
were eligible for enrolment. Patients were excluded when they had received exercise therapy in the past 3 months, a hip pain score <2 on an 11-point numeric rating scale (0 means no pain), a hip surgery in the past or were on the waiting list for hip surgery, severe disabling co-morbidity that disallowed receiving exercise therapy, insufficient comprehension of the Dutch language and/or were mentally incapable of participation. Written informed consent was obtained from all patients and the study was approved by the Erasmus MC Medical Ethical committee.
Using a computer-generated list, patients were randomised to exercise therapy added to GP care (intervention group) or to GP care only (control group) by an independent researcher who was blinded to the GP assigning patients. Patients in the intervention group were appointed to standardised exercise therapy. The exercise therapy, supervised by physiotherapists, consisted of (maximally) 12 evenly spread treatment sessions during the first 3 months followed by 3 booster sessions at 5, 7 and 9 months follow-up. In both groups, patients received unrestricted visits to their GP similar to a normal care situation: the GP provided education and counselling, prescribed pain medication if applicable, referred the patient to an orthopaedic surgeon and/or requested additional diagnostic examinations. As this study uses the intention to treat principle, a minority of patients received a hip surgery during follow-up. Enrolment commenced in September 2009 and finished in October 2011. The follow up period was 1 year.
The primary outcome measures of the randomised clinical trial included hip pain and hip-related activity limitations
. Details of these clinical results will be reported in a forthcoming publication. The present paper will focus on the cost-utility study.
The cost-utility study was primarily conducted from a societal perspective, but the healthcare perspective was also applied. Data on direct medical costs, productivity costs and quality of life (QoL) was collected using standardised questionnaires which were sent to the home addresses of the patients at baseline and at 6, 13, 26, 39 and 52 weeks follow-up. The recall period was either 6 weeks (at 6 and 13 weeks) or 13 weeks (at baseline, 26, 39 and 52 weeks). Annual costs were determined by adding up the costs per period. The costs for the time between the measurement periods (week 6–7) were established through linear interpolation. The naïve imputation strategy was used for missing values. All costs were based on Euro 2011 cost data. Where necessary, costs were adjusted to 2011 using the general price index from the Dutch Central Bureau of Statistics.
Direct medical costs
Total direct medical costs for individual patients were determined by multiplying resource use by corresponding unit prices. Data on resource use of visits to healthcare providers (GP, physiotherapist, medical specialist, company physician, psychotherapist and rehabilitation specialist), inpatient hospital days, rehabilitation center, nursing home and residential home, medical imaging (X-rays and magnetic resonance imaging), laboratory services, medications, appliances (cold and hot compresses, orthopaedic insoles and wheelchairs) and home care was acquired from the questionnaires. Additionally, patients could specify visits to ‘other professionals’, e.g., acupuncturist, masseur, aquatherapist. Data on hip surgeries were taken from the clinical study records. With respect to the intervention group, resource use of visits to the physiotherapist was additionally obtained from the physiotherapist.
Resource use of visits to healthcare providers, inpatient days, laboratory services and home care was valued using reference unit prices
. The resource use of medical imaging services was valued using fees as issued by the Dutch Healthcare Authority. Wholesale prices were used to value the resource use of medications and appliances. Cold and hot compresses were assumed to be used once monthly. Other appliances were assigned a life expectancy of 7 years.
The productivity costs involved productivity losses resulting from absence from paid work and reduced efficiency at paid and unpaid work. The number of absent days from paid work due to hip OA was valued using reference hour prices of productivity costs per paid employee, corrected for elasticity of labour time to avoid that differences in productivity losses between the intervention and control group would be caused by (income) differences related to chance
Reduced efficiency at paid work was also valued using the reference hour prices. The efficiency loss was established by means of the quality- and quantity method as developed by Brouwer et al. (1999) and incorporated in the PRODISQ instrument
. Patients gave their mark for the quality of their work on the last working day on a visual analog scale from 0 (worst quality) to 10 (best quality). The same question was posed for the quantity of their work on their last working day. These marks were assumed to be representative for the overall recall period. The efficiency loss during paid work in terms of hours lost was determined at (1−(quality/10) × (quantity/10)) × working hours per day.
Regarding unpaid work, patients were asked to indicate how many hours of housekeeping tasks were taken over by their family, other people and paid aid due to hip OA. These hours were valued using the current price of simple professional home care
QoL was measured by means of the EQ-5D instrument. The EQ-5D has five dimensions: mobility, self-care, activity, pain and anxiety. Each dimension has three levels: no problems (level 1), some problems (level 2) and serious problems (level 3). Hence, EQ-5D has 243 possible health states. Utility values for these health states were measured with the time trade-off technique on a random age-adjusted sample of the general adult population of the Netherlands
. All QoL measurements were analysed in one saturated linear model with correlated errors and an unstructured covariance matrix. This method for repeated measures takes the interdependence of observations within patients at different measurement moments into account, adjusts for differences at baseline and prevents bias caused by missing observations. Explanatory variables were the measurement moment and the interaction of treatment group and measurement moment after baseline. The interaction represented the estimated treatment effect at each moment. Regression results were used to predict mean QoL for both groups and for each measurement. Intrapolation of these predictions led to estimates of the mean number of QALYs per group.
Costs were averaged per cost category. The uncertainty round point estimates of incremental costs and effects were assessed by means of non-parametric bootstrapping. We drew 5000 bootstrap samples with replacement from the original sample. For each bootstrap sample, incremental costs and effects were estimated. 95%-confidence intervals (CI) were constructed by taking the 2.5 and 97.5th percentiles of the bootstrapping results for each outcome. The bootstrapping results were also used to present the combined uncertainty of incremental costs and effects graphically on CE-planes
. Each combination of incremental health effect and incremental cost was depicted as a dot on the plane. In addition, an acceptability curve was generated to indicate the probability that the intervention has lower incremental costs per QALY gained than various thresholds for the maximum willingness-to-pay for an extra QALY.
A total of 918 patients were invited for participation by the GPs during the recruitment period. Assessed for eligibility were 450 patients of which 247 did not meet our inclusion criteria. Thus, 203 patients were recruited, of which 101 were appointed to the intervention and 102 to the control group. For the intervention group, 100% of the questionnaires were returned at baseline, 91% after 6 weeks, 89% after 13, 26 and 39 weeks and 95% after 52 weeks. For the control group, 100% of the questionnaires were returned at baseline, 86% after 6 weeks, 78% after 13 weeks, 87% after 26 and 39 weeks and 91% after 52 weeks. The number of missing values was limited: 16 values in the intervention group (four GP visits, eight medical specialist visits, one inpatient hospital admission and three absences from work) and 12 values in the control group (two GP visits, four medical specialist visits, one inpatient hospital admission, five absences from work and two housekeeping tasks taken over).
Table I presents the general characteristics at baseline of the patients in the two groups. Two thirds of the patients were females. The mean age of the patients and the body mass index were slightly lower in the intervention than in the control group.
Table IGeneral characteristics of the included 203 patients at baseline
Annual direct medical costs for both the intervention and the control group are presented in Table II. The direct medical cost estimates were € 1233 (median 711) for the intervention and € 1331 (226) for the control group. Visits to the physiotherapist and hip surgeries were the key cost drivers. As intended, patients in the intervention group visited the physiotherapist more often.
Table IIAnnual mean direct medical costs (median) (Euro 2011)
Prior to the start of the study, about 3 out of 10 patients visited the GP and 1 out of 10 patients visited the physiotherapist. With respect to the physiotherapist, the fraction in the intervention group showed a fast increase to 90% at 6 weeks and a steady decrease from 13 weeks onwards to an average of 28% at 52 weeks. The average number of visits per patient visiting the physiotherapist was about 2 per month over time. In the control group, the fraction of patients showed a continuous pattern of about 19% during the entire follow up. The average number of visits per patient visiting the physiotherapist was about 3 per month, caused by a few patients with a relatively high number of visits.
Due to the high unit costs of hip surgery (€ 4754), hip surgeries were responsible for about a quarter of direct medical costs. Yet, only six patients in the intervention and nine in the control group received a hip surgery. The consumption of visits to ‘other professionals’, e.g., acupuncturist, masseur, aquatherapist, was negligible for both groups and none of the patients were admitted to a rehabilitation center, nursing or residential home. Medication was used by about 56% of the patients during follow up. In both groups, 46% of the medications was prescribed by a physician. Medications most frequently used were paracetamol, diclofenac, nurofen, etoricoxib, meloxicam and tramadol.
Table III presents productivity costs due to absence from paid work per measurement moment. Annual costs per patient were € 120 (SD 791) for the intervention and € 399 (SD 3008) for the control group. Slightly, but not significantly, more patients in the intervention group had a paid job. At most measurements, no patients in the intervention group were absent, while there were absent patients in the control group at any time. Still, the number of patients absent from work was limited to one or two patients, with the exception of the control group at 39 weeks (five patients).
Table IIIProductivity costs per measurement moment (Euro 2011)
Table III also presents efficiency losses at each measurement moment. The annual costs due to reduced efficiency at paid work were € 1215 (SD 3822) for the intervention and € 1396 (SD 5770) for the control group. Reduced efficiency of unpaid work was comparable between groups. Seventy-three percent of the patients in the intervention and 72% in the control group indicated that the reduced efficiency was caused by hip OA.
Regarding unpaid work, annual costs of taking over housekeeping tasks were € 66 (SD 160) for the intervention and € 114 (SD 324) for the control group. Paid aid incurred a minority of the costs (16% in the intervention and 21% in the control group).
Adjusted QoL scores on the EQ-5D were stable at 0.77 over time (compared to 0.87 in the general population
) and never significantly different between both groups. Over the full study period, the intervention and control group experienced the same amount of QALYs. The very small adjusted difference was 0.006 in favour of the control group (95%-CI: −0.04 to +0.02). Inspecting each EQ-5D dimension, the intervention group only had significantly less problems on the domain of daily activities at 6 and 13 weeks.
Table IV provides the total annual societal costs per patient in the intervention and control group. The total annual costs per patient were € 607 lower for the intervention group compared to the control group (95%-CI: −€ 2403 to +€ 993). This resulted in a societal average CE-ratio of € 97195 saved per QALY lost. However, the uncertainty around this CE-ratio was substantial. The CE-plane showed that the intervention had a 33% probability for positive health effects and 76% for cost savings (Fig. 1). The acceptability curve showed a probability of 68% that the intervention was cost-effective at a threshold of € 20,000 per QALY (Fig. 2). The curve slopes downward, due to the likelihood of small negative health effects; these effects appear less acceptable when more value is attached to health.
Table IVThe total annual costs per respondent in the intervention and the control group (SD)
When only direct medical costs were included, average incremental costs were € 98 lower per patient for the intervention group and average savings per QALY lost were € 15,692. The intervention had a 33% probability for positive health effects and 61% for cost savings. The acceptability curve showed a probability of 47% that exercise therapy was cost-effective at a threshold of € 20,000 per QALY.
This is the first economic evaluation on exercise therapy added to GP care from a societal perspective in patients with hip OA in primary care. Annual direct medical costs per patient were significantly lower for the intervention (€ 1233; SD 1864) compared to the control group (€ 1331; SD 2773) despite additional physiotherapy visits. Productivity costs amounted to € 1401 (SD 3976) and € 1910 (SD 6429) for the two groups respectively. Thus, annual societal costs per patient were considerably lower for the intervention group (€ 2634 vs € 3241), which confirms that the inclusion of productivity costs may affect total costs and CE-ratio.
When costs are determined from a societal perspective, all relevant costs should be included
. However, in some jurisdictions only direct medical costs are required in economic evaluations (e.g., the United Kingdom). As expected, our results suggest that productivity costs are more important than direct medical costs. Particularly costs which occurred due to reduced efficiency at paid work were substantial. Hence, productivity cost reductions might partially compensate for the additional cost of exercise therapy.
Only 90% of the patients in the intervention group visited a physiotherapist at 6 weeks. This fraction showed a steady decrease to an average of 28% at 52 weeks. This implies that some of the intervention patients did not meet the terms of the standardised exercise program they were appointed to. Assuming that the patients have gained all the benefit they need by 13 weeks, the efficiency and CE of the intervention could even be better in case of a program with only 3 months physiotherapy. On the other hand, patients may not have been able to sustain treatment, despite the benefit of longer treatment. In this scenario, there seems to be room for actions to improve compliance, which could make the program more (cost-)effective.
Remarkably, 13 patients in the intervention group reported zero visits, although the physiotherapist did record visits. In these cases, the number of visits as provided by the physiotherapist was used. Additionally, only 11% of the patients reported exactly the same number of physiotherapy visits as the physiotherapist. Of the remaining patients, 52% reported less and 38% more visits per year than the physiotherapist. One-way sensitivity analyses altering physiotherapy costs between 50% and 150% revealed that direct medical costs varied ±24% for the intervention and ±8% for the control group. Societal costs varied ±11% for the intervention and ±3% for the control group. The average numbers of visits per year were 11.8 according to the patients and 8.6 according to the physiotherapists, which was lower than the projected maximum of 15.0 visits. Even though the use of two independent sources for the cost calculation generally provokes inconsistency, it takes advantage of more complete data.
Although having had a hip surgery in the past or being on the waiting list for hip surgery was an exclusion criterion, a minority of patients received a hip surgery. Hip surgeries were responsible for about 25% of direct medical costs. In addition, patients receiving hip surgery had relatively high physiotherapy costs (€ 672 in the intervention and € 1062 in the control group). An additional analysis excluding patients receiving hip surgery revealed direct medical cost estimates for the intervention group to be € 804 (versus € 1233 including these patients) and for the control group € 537 (versus € 1331). This indicates that hip surgeries may be considered as an effect modifier. However, as this is a pragmatic trial using the intention to treat principle, these surgeries should be regarded in the direct medical costs. After all, only correcting for costs biases the results, since hip surgeries not only affect costs, but also QoL, functioning and pain.
Our study concerned the inclusion of a small number of patients. The sample size was based on power calculations for the primary outcome measures of the randomised controlled trial in conjunction with which our cost-utility study was performed (hip pain and hip-related activity)
. Costs generally have a greater variation and skewness than clinical outcome measures. Power calculations for the CE-ratio would probably have indicated the requirement of a much larger sample size which would not have been fundable in the Dutch context.
To prevent any bias in QALY estimates, we adjusted for possible imbalance in baseline utility
. This resulted in a very small adjusted difference in QoL of 0.006 in favour of the control group. The 95%-CI is close to zero (−0.04 to +0.02), which indicates that exercise therapy is unlikely to lead to noteworthy negative health effects. Furthermore, QoL was fairly stable at 0.77 over time and never significantly different between the intervention and control group. This finding is in agreement with the study of Cochrane et al. (2005), who evaluated the CE of water exercise over usual care in the management of lower limb OA. The average difference in effects was found to be only 0.013
Our study revealed that exercise therapy is probably cost saving, without the risk of noteworthy negative health effects. There seems to be a rationale to question current usual care, but an efficient policy concerning physiotherapy requires treatment consensus and optimal interaction with other health providers such as GPs and medical specialists. Furthermore, it requires convincing evidence of the CE of exercise therapy over GP care alone. Therefore, future studies should investigate whether continued exercise therapy would raise health effects and improve the CE-ratio. Our uncertainty analysis indicated that there is a probability of 33% that exercise therapy produces positive health effects and 76% that exercise therapy saves societal costs compared to GP care alone. Whether our results are sufficiently acceptable to use exercise therapy additional to GP care is up to the decision maker (e.g., policy maker, GP or patient).
This study was conducted in the Netherlands. However we believe that our resource use findings could be representative of other countries, especially those in which the GP operates as the gatekeeper of healthcare.
The project was coordinated by PAJL and MAK. SST drafted this article, together with CHT and PPvE. LMAG was responsible for the statistical soundness of the study. The conception and design of the study, acquisition of data, analysis and interpretation of data were all done jointly by all authors. JD, AMB, JANV, BWK, SMAB-Z, PAJL and MAK all revised the draft article critically for important intellectual content and approved of the version to be submitted.
Role of the funding source
This research was financial supported by the Netherlands Organisation for Health Research and Development (grant 170992402 ). The study sponsor did not have any role in the study design, collection, analysis and interpretation of data; in the writing of the manuscript; and in the decision to submit the manuscript for publication.
Competing interest statement
None of the authors has any financial or personal relationships with other people or organizations that could potentially and inappropriately influence (bias) their work and conclusions.
We like to thank Mrs. T. Mulder – van Kempen and Ms. B. Schutijser for collecting the data of the trial.
Rijksinstituut voor Volksgezondheid en Milieu (RIVM)
Nationaal kompas volksgezondheid (National Public Health Compass).