|
A cost-effectiveness analysis of two rehabilitation support services for women with breast cancer |
Gordon L G, Scuffham P, Battistutta D, Graves N, Tweeddale M, Newman B |
|
|
Record Status This is a critical abstract of an economic evaluation that meets the criteria for inclusion on NHS EED. Each abstract contains a brief summary of the methods, the results and conclusions followed by a detailed critical assessment on the reliability of the study and the conclusions drawn. Health technology The study examined two rehabilitation interventions for breast cancer survivors. The first intervention, Domiciliary Allied Health and Acute Care Rehabilitation Team (DAART), was described as an early, home-based physiotherapy intervention. The second intervention, Strength Through Recreation Exercise Togetherness Care Health (STRETCH), was described as a group-based exercise and psychosocial intervention. Both interventions were described in full detail in the pape
Study population The study population comprised women aged between 25 and 74 years who were diagnosed with primary breast cancer. The inclusion criteria were that the women had unilateral disease, spoke English and had no cognitive problems. Women were excluded if they were considered too ill or had previously received either intervention.
Setting The setting was community care. The economic study was carried out in Australia.
Dates to which data relate The effectiveness data were collected from women recruited between May 2002 and July 2003. The cost data referred to the 12 months prior to study recruitment. The price year was 2004.
Source of effectiveness data The effectiveness data were derived from a single study.
Link between effectiveness and cost data The cost data were collected prospectively from the same patient sample as that used in the effectiveness study.
Study sample The authors did not report any power calculations. They assembled a convenience sample by inviting consecutive patients who routinely attended DAART or STRETCH to participate in the study. For the non-intervention group, patients were sourced from a concurrent breast cancer research project. The authors stated that the comparison group was selected to be representative of women with breast cancer in the same geographic area as the women receiving the intervention. The authors stated that there were some important baseline differences between the three patient groups (two intervention and one non-intervention) in terms of their socio-demographic profiles, severity of disease, and treatment outcomes and general health. The authors did not report any detail about the number of eligible women identified or the number who refused to participate. However, they did state in the discussion that fewer eligible women were recruited than anticipated. The study sample comprised 36 women who received DAART, 31 who received STRETCH, and 208 who received no intervention.
Study design This was a prospective cohort study. Patients were recruited from multiple centres and followed up to 12 months from the date of their diagnosis. The authors did not quote any loss to follow-up. It is possible that they conducted a complete case analysis.
Analysis of effectiveness It was difficult to ascertain from the data reported whether all patients included in the study were accounted for in the analysis. The primary health outcomes used were utility scores and number of rehabilitated cases. The utility scores were measured using a single-item linear analogue scale, but the authors did not state at which time point they were measured. Rehabilitated cases were defined according to the Functional Assessment of Cancer Therapy - Breast Cancer plus arm morbidity scale (FACT-B+4), with improvements measured from 6 to 12 months post-diagnosis. Several confounding variables were identified, such as country of birth, level of education, income, presence of health insurance, time since diagnosis at pre-intervention, treatment received and tumour size. The authors reported that the utility scores were adjusted for potential confounders.
Effectiveness results Patients receiving DAART had an average adjusted utility score of 0.84 (95% confidence interval, CI: 0.77 to 0.90).
Patients receiving STRETCH had an average adjusted utility score of 0.80 (95% CI: 0.73 to 0.87), while the non-intervention group had an average adjusted utility score of 0.72 (95% CI: 0.70 to 0.75). The differences between the groups, in terms of the utility scores, were found to be significant, (p=0.003).
A chi-squared test found no significant differences between the groups in terms of the proportion of rehabilitated cases, (p=0.80).
Clinical conclusions The authors concluded that the outcome of rehabilitated cases measures a different concept to the outcome measure of utility. They conclude that the choice of outcome measure will have an important impact on the interpretation of the study results.
Modelling A decision tree model was used to calculate the cost per quality-adjusted life-year (QALY) for each of the treatment strategies. The authors stated that the model was analysed using second-order Monte Carlo simulation. However, the subsequent description of the methods used implied a first-order or patient-level simulation. Taking a Bayesian approach, the model was used to calculate the probability that each intervention would be cost-effective for a particular patient population.
Measure of benefits used in the economic analysis The study relied on two measures of benefits for the economic analysis. The utility scores were multiplied by one year to generate QALYs. The effectiveness outcome of the proportion of rehabilitated cases was also used.
Direct costs The costs and the resource quantities were not reported separately. The study included costs to the health service, patients and societal costs. The direct costs covered personnel, capital equipment, consumables, operating expenses, travel, parking and childcare. The resource use quantities were obtained from patients using a retrospective questionnaire. The authors did not describe the unit costs in sufficient detail. Discounting was not relevant given the one year timeframe for the analysis. The study reported the average costs. All costs were converted to 2004 prices using the all items Consumer Price Index (CPI).
Statistical analysis of costs The costs were not analysed statistically, but they were used to inform a probabilistic decision model. Thus, the underlying distribution of the costs was used to inform the distribution used in the decision model. The authors reported the mean costs with standard deviations.
Indirect Costs The indirect costs were included to provide the study with a broad societal perspective. The study included the costs to the community (lost productivity and volunteer time to run the interventions) and the costs to the patients (leisure time forfeited and time spent on health care). The value of volunteer time was measured using the market replacement cost method, as was unpaid work. Paid work was valued using Census data. Details about lost productivity, work and leisure time were obtained from patient questionnaires. Leisure time and time spent attending health care sessions were valued using average wages. The quantities and the costs were not analysed separately. The indirect costs were measured over one year, so discounting was not relevant. The price year was 2004. Adjustments for inflation were made using the all items CPI.
Currency Australian dollars (AUD).
Sensitivity analysis The underlying distributions in the sampled data were used to inform distributions in the decision model. The authors investigated variability in the data in a probabilistic sensitivity analysis. The authors stated that a second-order Monte Carlo simulation was performed, but their description implied a first-order Monte Carlo simulation. Several one-way sensitivity analyses were also performed to check how robust the model results were to extreme values and to measure the effect of uncertainty.
Estimated benefits used in the economic analysis The authors reported the incremental QALYs gained in the Monte Carlo simulation of the decision model. DAART was estimated to result in a gain of 0.l00 QALYs compared with no intervention over a period of one year. STRETCH was estimated to result in a gain of 0.067 QALYs compared with no intervention over a period of one year.
Cost results DAART was estimated to cost AUD 129 more than no intervention in a probabilistic analysis using rehabilitated cases as the outcome measure. The corresponding cost estimated in a probabilistic analysis using QALYs as the measure of economic benefit was AUD 136.
STRETCH was estimated to cost AUD 936 more than no intervention in a probabilistic analysis that used rehabilitated cases as the outcome measure. The corresponding cost in the analysis that used QALYs was AUD 940.
Synthesis of costs and benefits The costs and benefits were synthesised to calculate the cost per rehabilitated case and the cost per QALY gained.
Both interventions were dominated by no intervention when the outcome measure was rehabilitated cases. DAART was estimated to cost AUD 1,351 (95% CI: -18,086 to 19,516) per QALY gained compared with no intervention.
STRETCH was estimated to cost AUD 13,989 (95% CI: -77,826 to 47,935) per QALY gained compared with no intervention.
The 95% CIs for the incremental cost-effectiveness ratios (ICERs) were obtained by rank ordering the ICERs from the probabilistic analysis and identifying the upper and lower 2.5th percentile. This is not an appropriate method for describing the uncertainty around ICERs because of the difficulty in interpreting negative ratios. However, the study also provided cost-effectiveness acceptability curves, which are an appropriate way to characterise second-order uncertainty although they appear to have been used to describe first-order uncertainty.
Authors' conclusions The Domiciliary Allied Health and Acute Care Rehabilitation Team (DAART) intervention is the most cost-effective treatment option for providing rehabilitative care for breast cancer.
CRD COMMENTARY - Selection of comparators The justification for the comparators used was that they represented current practice in the study setting. You should decide whether these are widely used health technologies in your own setting.
Validity of estimate of measure of effectiveness The effectiveness data were derived from a single cohort study. The study design was not appropriate for the study question as the groups were not comparable at baseline and the sample suffered from selection bias. Although the authors stated that they attempted to control for confounding factors, the methods they used were unclear and the costs were unadjusted in the main analysis. In addition, the outcome measure of proportion of rehabilitated cases does not appear to have been based on adjusted analysis. The study sample for the non-intervention group was representative of the study population, but the patient samples in the intervention groups might not have been. The authors did not compare the characteristics of those women that agreed to participate in the study with those that refused. These factors limit the validity of the study results.
Validity of estimate of measure of benefit Two measures of benefit were used in the economic analysis. The estimation of benefits was based on a decision tree model. The parameters for the decision model were derived from the effectiveness study. The use of QALYs as a measure of economic benefit is difficult to interpret as the authors did not report the time at which utility was measured. The way in which the decision model was used to calculate QALYs and rehabilitated cases was not described in sufficient detail. These factors limit the validity of the study results.
Validity of estimate of costs The authors included all costs relevant to the societal perspective adopted. For each category of cost all relevant costs were included. The costs were not reported separately from the quantities. Although the study was stated as prospective, the cost data were measured by patient questionnaire that relied on patient recall. The resource use quantities were not analysed statistically, despite the availability of patient-level data. It was unclear whether the resource use estimates were adjusted for multiple confounding factors. The underlying distribution of the costs was, however, used to inform distributions in a probabilistic sensitivity analysis. The authors did not describe the source of the unit cost data for the direct costs, which limits the generalisability of the study results. The authors did report the source of the unit costs for the indirect cost items, but they did not report the actual unit costs. This also limits the generalisability of the study results.
Other issues The authors did not compare their findings with those from other studies, stating instead that no such analysis had been performed previously in the study setting. The issue of generalisability to other settings was not addressed. The authors do not appear to have presented their results selectively, but the poor reporting of some of the methods used makes the results difficult to interpret. The authors' conclusions did not reflect the scope of the analysis as they attempted to compare DAART and STRETCH on the basis of two separate pair-wise comparisons with no intervention. It would have been more appropriate to simultaneously compare DAART, STRETCH and no intervention in the same analysis. The authors stated that the failure to recruit as many patients as anticipated for the intervention programmes, and the likely imbalance of those groups as a result of the poor recruitment, are limitations of the study. They also acknowledged the ambiguous nature of the study results depending on which of the two outcome measures was selected.
Implications of the study The authors recommended that both interventions should be considered good value.
Source of funding Supported by the National Breast Cancer Foundation and Women in Super.
Bibliographic details Gordon L G, Scuffham P, Battistutta D, Graves N, Tweeddale M, Newman B. A cost-effectiveness analysis of two rehabilitation support services for women with breast cancer. Breast Cancer Research and Treatment 2005; 94(2): 123-133 Other publications of related interest Gordon LG, Battistutta D, Scuffham PA, et al. The impact of rehabilitation services on health-related quality of life for women with breast cancer. Breast Cancer Res Treat 2005 (in press).
Indexing Status Subject indexing assigned by NLM MeSH Breast Neoplasms /rehabilitation; Community Health Services /economics; Cost-Benefit Analysis; Exercise Therapy /economics; Female; Home Care Services; Humans; Middle Aged; Outcome Assessment (Health Care); Physical Therapy Modalities /economics; Quality-Adjusted Life Years; Queensland; Survivors AccessionNumber 22005001784 Date bibliographic record published 30/09/2006 Date abstract record published 30/09/2006 |
|
|
|