Home > Full Text Reviews > Hemiarthroplasty and Total Hip... > Assessment of cost-effectiveness

PubMed Health. A service of the National Library of Medicine, National Institutes of Health.

Carroll C, Stevenson M, Scope A, et al. Hemiarthroplasty and Total Hip Arthroplasty for Treating Primary Intracapsular Fracture of the Hip: A Systematic Review and Cost-Effectiveness Analysis. Southampton (UK): NIHR Journals Library; 2011 Oct. (Health Technology Assessment, No. 15.36.)

4Assessment of cost-effectiveness

Methods for reviewing cost-effectiveness

A review of the evidence for cost-effectiveness has also been undertaken. The searches performed were as described in Chapter 3, Identification of studies, but slightly different study selection criteria were applied to the results. Studies with either the outcomes of resource utilisation or cost–utility (as listed in Chapter 3, Secondary outcomes) or economic evaluations relating to the population and interventions specified in Chapter 3, Inclusion criteria, were included.

Results

Quantity of research available

The search of electronic databases identified 532 unique citations. Seven full papers were retrieved to determine whether or not they were relevant to this review. After screening, four studies satisfied the inclusion criteria.31,4446 Details of the screening and inclusion process are provided in the PRISMA flow chart (Figure 10). Three studies used mathematical models to perform an economic evaluation,4446 and one paper31 reported the costs and utilities collected alongside an RCT.

FIGURE 10. PRISMA flow diagram – cost-effectiveness.

FIGURE 10

PRISMA flow diagram – cost-effectiveness.

A review of the cost-effectiveness literature

Four papers were identified as having an economic element, although only one took the form of a cost–utility analysis. This paper46 evaluated a patient population with a displaced femoral neck fracture who were elderly and active and treated in an American setting. All costs associated with the surgical procedure and future revisions were included in a Markov model. The conclusion from the model was that THA was the more cost-effective treatment in the patient population with an expected 1.53 quality-adjusted life-years (QALYs) being provided at a cost of US$3000. The cost per QALY ratio of US$1960 would be viewed as extremely cost-effective using standard UK cost-effectiveness thresholds.47 It was seen that the key driver of this result was the increased utility associated with patients who had undergone THA compared with those that had HA. These data were taken from Keating et al.,31 with the difference in utility shown to be significant at 24 months (p = 0.008).

The study by Aleem et al.45 did not include the costs associated with either surgical procedure and focused on the procedure that produced the greatest patient benefit, which implicitly assumes that costs are equal for THA and HA. The derivation of the utilities used within the model was far from ideal as these were derived from asking surgeons and hypothetical patients to rate model outcomes in terms of 0 (death) to 100 (perfect health), rather than using utilities reported directly from patients and with the derived utilities based on public preferences, using a choice-based method, as recommended by the National Institute for Health and Clinical Excellence (NICE).47 Additionally, median values were used rather than mean values, which is incorrect in economic evaluations. The authors concluded that arthroplasty produces better patient outcomes than internal fixation and that THA had slightly better outcomes than HA.

The analysis of Iorio et al.44 reported outcomes in terms of the cost per ambulatory patient at 2 years for four procedures (reduction with internal fixation, unipolar HA, bipolar HA and THA). As such, differences in the quality and length of life of patients during this period were ignored and the paper in essence reports a cost minimisation analysis. The authors concluded that THA was the most cost-effective of the four procedures.

Keating et al.31 reported the costs and utility consequences from an RCT which compared THA and HA. The authors claim that THA is more cost-effective, but do not provide incremental cost-effectiveness ratios. The data contained within this manuscript could be used to calculate an estimate of the likely cost-effectiveness of THA compared with HA at the duration of follow-up (2 years) and at extrapolated time horizons. The authors of this report undertook this using a simple mathematical model that is detailed later (see Chapter 4, The economic evaluation undertaken within this report).

An assessment of Slover et al.,46 using the Drummond et al.48 checklist, is contained in Appendix 4. The remaining three papers with economic elements were not assessed as they were considered less appropriate owing to undertaking either a cost minimisation44 or a benefit maximisation45 approach or simply reported data from an RCT.31

The economic evaluation undertaken within this report

It was deemed that the Iorio et al.44 and Aleem et al.45 studies were too limited to inform the decision problem fully. The paper by Slover et al.46 was a mathematical model of reasonable quality, but was based on a US setting rather than a UK one. The authors of this report decided to perform an economic evaluation based on the Keating et al.31 RCT as this had high internal validity and was directly applicable to the study population. If the results from this analysis concurred with those from Slover et al.,46 and to a lesser extent those of Iorio et al.44 and Aleem et al.,45 then this would support the conclusions that THA was more cost-effective than HA. The proposed modelling methodology was discussed with the clinical expert, who deemed that this was an acceptable conceptual model. Given the resource constraints, a decision to employ a simplistic model was undertaken.

The Keating et al.31 RCT is directly relevant to the decision problem as it was conducted in Scotland and compared the two interventions of interest. The data reported contained the utility of patients at 4, 12 and 24 months using the EQ-5D questionnaire and the mean costs associated with each intervention over the 2-year period. The EQ-5D is the utility measure preferred by NICE.47 Costs were presented in five categories: initial inpatient episode; hip-related admissions; non-hip-related admissions; total hip-related costs; and total costs. Data concerning the characteristics of the Keating et al.31 RCT are presented in Table 1. Owing to the direct relevance and high internal validity, the authors believed that these data were more appropriate to populate the economic model than the results produced by the meta-analyses undertaken earlier in this report.

An estimate of the cost-effectiveness of THA compared with HA was calculated assuming that the increased costs associated with THA were normally distributed with a mean of £3010 with a standard error (SE) of £2250. This cost differential is given some support by data from the American 2003 National Inpatient Survey reported in Slover et al.46 that stated that the average hospital charges for THA compared with HA were US$4409 higher. The costs from Keating et al.31 were inflated from the 2000–1 price year to a 2007–8 price year,49 resulting in a mean increase in costs associated with THA compared with HA of £3937; the SE of this increase was assumed to increase to £2943. It was assumed that all costs were incurred in the first year and that costs would remain constant for both arms for the remainder of the model. This approach has support in research undertaken by Haentjens et al.,50 which indicated that the type of surgical procedure (THA or HA) was not associated with differential costs in the year following hospital discharge. Given this methodology, costs were not discounted.

Based on distributions presented in Keating et al.,31 it was assumed that the EQ-5D increase was 0.09 (SE 0.05), 0.05 (SE 0.05) and 0.16 (SE 0.06) at 4, 12 and 24 months, respectively. It was assumed that there was a linear change from zero to the sampled difference in utility at 4 months, a linear change between the sampled differences at 4 and 12 months and a linear change between the sampled differences at 12 and 24 months. The difference at 24 months was assumed to persist until the end of the modelling horizon. Utilities were discounted at 3.5% per annum as recommended by NICE.47 In the analyses undertaken, time horizons of 2, 3 and 5 years were assessed as it was believed that the vast majority of patients who were alive at 2 years would survival an additional 3 years.

The incremental cost per QALY of THA was calculated as the incremental cost of THA divided by the incremental QALY. A plot of the modelled utilities is provided in Figure 11 assuming that the midpoint estimates for both THA and HA are correct.

FIGURE 11. The assumed gain in utility associated with THA compared with HA.

FIGURE 11

The assumed gain in utility associated with THA compared with HA.

The mortality rates observed within the trial were considered. In the Keating et al.31 RCT there was a greater proportion of deaths in the HA arm (13%) than in the THA arm (9%), although this was not statistically significant (p = 0.36). These data were pooled to form a risk of mortality in both arms of 11%, and it was assumed that the incremental QALY gain estimated for THA would be reduced by 11% to account for mortality.

In order to preserve consistency between the sampled utility differences when conducting the probabilistic sensitivity analyses, the same random number was used to select from the cumulative distribution function for each time point. This would ensure that if the value sampled for the difference at 4 months was higher than the median; the differences at 12 and 24 months would also be higher than the median value.

For clarity, the parameter values used in the probabilistic sensitivity analyses are given in Table 10.

TABLE 10. The parameters used within the economic model.

TABLE 10

The parameters used within the economic model.

The results from this model are provided in Table 11 and used 1000 Monte Carlo simulations. It is seen that even when the utility benefits are constrained to the 2-year horizon the cost per QALY is < £30,000. When the time horizon is extrapolated to more realistic values, the cost per QALY decreases, reaching a value < £10,000 with a horizon of only 5 years. This value would be seen as cost-effective under current cost-effectiveness thresholds.47 It is seen that the results produced within our analyses concur with previous authors4446 in that THA is likely to be more cost-effective than HA.

TABLE 11. The results from the model when comparing THA with HA.

TABLE 11

The results from the model when comparing THA with HA.

The likelihood of THA being more cost-effective than HA can be displayed on a cost-effectiveness acceptability curve; this is shown in Figure 12. All time horizons are shown simultaneously on this figure for brevity; these are different modelling scenarios rather than competing strategies within one decision problem.

FIGURE 12. Cost-effectiveness acceptability curves depicting the likelihood that THA is more cost-effective than HA.

FIGURE 12

Cost-effectiveness acceptability curves depicting the likelihood that THA is more cost-effective than HA.

Limitations of the analyses

It is commented that longer-term consequences, such as the rates of revision and dislocation, have not been considered in this analysis. Data from studies with a follow-up to 13 years indicate that THA is associated with significantly fewer revisions (RR 0.31, 95% CI 0.17 to 0.59; see Table 3), whereas HA is associated with significantly fewer dislocations (RR 2.40, 95% CI 1.21 to 4.76; see Table 3). The impact of these omissions is likely to be unfavourable to THA as clinical advice indicates that the costs and disutility associated with revisions are far greater than those associated with dislocations. As such, this strengthens the conclusions that THA is more cost-effective than HA.

The effect of ageing on the incremental gain in utility has not been considered. There are no data to indicate whether or not the gain would increase, decrease or remain static as patients age; however, it is expected that the results may be more uncertain than presented.

Exploratory sensitivity analyses

Exploratory sensitivity analyses were undertaken assuming that the increased utility associated with THA compared with HA was equal to the midpoint reported by Blomfeldt et al.,32 which was 0.05 (0.68 for THA and 0.63 for HA). Although this difference was statistically non-significant, the indication from Keating et al.31 is that there is a real difference in utility. In this sensitivity analysis, the cost per QALY was £44,997, £30,511 and £18,932 at 2, 3 and 5 years, respectively. These values were was not as favourable to THA as the analyses based on Keating et al.,31 but they still indicate that THA is likely to be more cost-effective than HA assuming a time horizon of ≥ 5 years using standard UK thresholds.48

However, the authors prefer the data from Keating et al.31 as this study has a UK setting, has a slightly larger sample size, has a greater follow-up period and is consistent with the values used for increased costs associated with THA.

The cost data from Keating et al.31 were inflated to 2007–8 prices, and it is uncertain whether or not the costs originally reported would have risen equally for both THA and HA, although it is likely there would have been some correlation regarding costs, such as inpatient costs that would be incurred in each operation.

Additionally, although there is some support for equal costs after the 2-year period,50 it may be that the different rates of revisions and dislocations have a cost implication.

In order to explore the possibility that these incremental costs may differ from that used in the base case, sensitivity analyses were conducted varying the incremental cost of THA compared with HA (Table 12).

TABLE 12. Sensitivity analyses exploring the impact of different assumed incremental costs of THA compared with HA on the cost per QALY gained (£).

TABLE 12

Sensitivity analyses exploring the impact of different assumed incremental costs of THA compared with HA on the cost per QALY gained (£).

These sensitivity analyses indicate that even if the incremental cost of THA compared with HA increased to £8000 then it is likely that THA would still be cost-effective provided that the time horizon was ≥ 5 years, given current cost-effectiveness thresholds.47

The expected value of perfect information

The expected value of perfect information (EVPI)51 was calculated for the base-case model, assuming that the funders were prepared to pay a cost of £20,000 per QALY gained.47 Population EVPI provides the maximum that a funder would be prepared to pay to eliminate all uncertainty in the decision problem and thus know with certainty which option was more cost-effective. If the cost of the research required to provide further information is greater than the population EVPI then the research should not be funded.

The estimated EVPI per patient is given in Table 13 using time horizons of 3 and 5 years. At these time points the adoption decision would be THA. Population EVPI is calculated by the number of patients who are assumed to benefit owing to the greater certainty of which procedure is the more cost-effective.

TABLE 13. The EVPI per patient at different modelling time horizons.

TABLE 13

The EVPI per patient at different modelling time horizons.

As previously discussed, the omission of the costs and disutilities associated with revisions and dislocations is likely to strengthen the conclusion that THA is more cost-effective than HA. This would reduce the uncertainty in the decision and therefore it is likely that the EVPI is overestimated.

It is seen that the EVPI decreases as the modelling horizon increases. This is due to the greater certainty that THA is more cost-effective than HA when the time horizon is of larger duration. These values, however, are likely to change when trials currently under way report their findings and the evidence base expands.

© 2011, Crown Copyright.

Included under terms of UK Non-commercial Government License.

Cover of Hemiarthroplasty and Total Hip Arthroplasty for Treating Primary Intracapsular Fracture of the Hip: A Systematic Review and Cost-Effectiveness Analysis
Hemiarthroplasty and Total Hip Arthroplasty for Treating Primary Intracapsular Fracture of the Hip: A Systematic Review and Cost-Effectiveness Analysis.
Health Technology Assessment, No. 15.36.
Carroll C, Stevenson M, Scope A, et al.
Southampton (UK): NIHR Journals Library; 2011 Oct.

PubMed Health Blog...

read all...

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...