The FOREMOST trial
We have recently described the methodology of the FOREMOST trial, including inclusion and exclusion criteria and procedures for participants in each of the two study groups when non-reassuring fetal status was identified.8 Briefly, with ethical approval from the relevant ethics committees, the FOREMOST trial was conducted in four Australian maternity hospitals, each with more than 4000 births/year: the Royal Brisbane and Women’s Hospital, Brisbane, Queensland; the Mater Mothers’ Hospital, Brisbane, Queensland; the Royal Hospital for Women, Sydney, New South Wales and the Royal Women’s Hospital, Melbourne, Victoria, between July 1999 and September 2004. Consenting women whose fetuses showed a non-reassuring fetal heart rate pattern during labour were randomised to one of the two study groups: continued conventional CTG monitoring (control, CTG-only group) or the addition of FPO to the CTG monitoring (intervention, FPO + CTG group). In the FPO + CTG group, a sensor was placed on the fetal temple or cheek during a vaginal examination and connected to an oximeter (Nellcor OxiFirst; TYCO Inc., Pleasanton, CA, USA). There was a statistically significant 23% relative risk reduction in the primary outcome: operative delivery for non-reassuring fetal status in those allocated to the intervention group (FPO + CTG monitoring, n= 76, 25%) compared with those allocated to the control group (CTG only, n= 95 [32%]) (FPO + CTG : CTG only—relative risk 0.77, 95% CI 0.599–0.999, P=0.048, absolute risk reduction 0.07). Analyses in this study are based on the absolute risk reduction of 7%. There were no between-group differences in overall operative delivery rates or neonatal outcomes.8
Economic evaluation of the FOREMOST trial
The economic analysis of the FOREMOST trial was undertaken in Australian dollars, with the base year being 2004. Readers from other countries may wish to convert our findings to local currencies: currency conversion rates at June 2004 were $A1.00 equivalent to $US0.69030 or 0.57135 Euros. We used cost-effectiveness analysis (CEA) that compared the cost and effectiveness of the new clinical strategy: adding FPO to conventional CTG monitoring versus conventional CTG monitoring alone. This method determines the ‘price’ of the additional outcome purchased by changing from current practice (CTG only) to the new strategy (FPO + CTG). The CEA examined change in resource use and outcome at the margin for the prespecified primary outcome: operative delivery for non-reassuring fetal status averted by the addition of FPO. Costs were for treatment-related expenses and included direct medical costs, such as fetal monitoring procedures, equipment, medications, delivery and postnatal stay. A social perspective, incorporating all costs that may accrue, is the preferred viewpoint in economic evaluation. However, given the uniformity of the treatment environment in which the study was undertaken, we considered a limited perspective of costs from the viewpoint of the health service appropriate for this evaluation.13
Resources used in the provision of care were predominantly measured using a top-down approach and supplemented with bottom-up estimates. Clinical records provided information on the type and nature of delivery, specifically operative delivery for non-reassuring fetal status and overall caesarean section. Case notes were matched with relevant hospital-assigned diagnosis-related group (DRG) codes to estimate costs associated with delivery and neonatal admission in each hospital. DRGs are used in Australian hospitals to categorise patients according to the nature and type of care provided in the hospital.14 Such a system is preferred to the use of International Classification Disease (ICD) codes alone as it incorporates about 10 000 ICD codes into a manageable number of broad categories. All DRG costs were adjusted for inflation to costs in 2004 using the Consumer Price Index (CPI) index for hospital and medical services.15Table 1 shows the range of maternal delivery DRGs as an example. Details of the numerous possible neonatal DRGs are available on the Foremost website.16 Although DRGs are presented as average costs per episode of care, their derivation is based on a summation of direct and indirect costs calculated across a range of cost buckets (components of costs) divided by the total number of separations for that DRG. Cost buckets used in estimates included clinical and nonclinical services (ward medical and ward nonmedical), pathology, imaging, allied, pharmacy, critical care, operating rooms, emergency room departments, hotel and depreciation.17 For example, DRG O01D is the code for a caesarean delivery without complicating diagnosis. In New South Wales, DRG O01D costs $5353, comprising $3653 in direct costs and $1699 in overhead costs. The cost bucket responsible for the greatest resource use in this code is nursing, with direct costs of $1635 and overhead costs of $102. Two features of Table 1 are worth noting. First, cost weights vary by state and across hospitals within the same state, predominantly arising through variations in state-based hospital pricing arrangements. Second, operative deliveries are generally more expensive than vaginal deliveries, with costs varying in accordance with severity of complications.
Table 1. DRGs for birthing procedures in Australian hospitals17*
|DRG code||DRG description||Average cost per DRG ($A)|
|New South Wales||Victoria||Queensland|
|O01A||Caesarean delivery (multiple complicating diagnoses, at least one severe)||9236||13 171||6955|
|O01B||Caesarean delivery (severe complicating diagnosis)||6195||7742||5808|
|O01C||Caesarean delivery (moderate complicating diagnosis)||5618||6966||5830|
|O01D||Caesarean delivery without complicating diagnosis||5353||5786||5009|
|O02Z||Vaginal delivery with complicating operating room procedure||5478||5761||4641|
|O60A||Vaginal delivery (multiple complicating diagnoses, at least one severe)||4629||5477||3844|
|O60B||Vaginal delivery (severe complicating diagnosis)||3644||3824||3167|
|O60C||Vaginal delivery (moderate complicating diagnosis)||3581||3612||3225|
|O60D||Vaginal delivery without complicating diagnosis||3090||3005||2422|
The bottom-up approach to costing involved identifying resource use at the patient level specific to the trial and not otherwise captured within the DRGs: fetal scalp blood sampling, medications, use of oxygen and/or intravenous hydration fluid and intervention-related equipment including the FPO unit, FPO sensors, fetal blood sampling kits and fetal blood sampling analysis units.
To ensure that resource use data could be legitimately compared across the four trial sites, a standardised methodology was applied, which involved the application of uniform data collection procedures and valuation techniques. Resource use was predominantly ascertained using clinical records and specific data collected as part of the trial. Resources were valued according to local prices. Medications were valued using the Australian Pharmaceutical Benefit Scheme.18 The annuitised cost of intervention-specific equipment was valued from the source.
Consistent with guidelines for conducting economic evaluation,19 results are presented in terms of an incremental cost-effectiveness ratio (ICER) to ascertain the additional cost per unit of benefit associated with moving to the new assessment technique. The ICER is calculated using the following formula:
where i is the intervention group (FPO + CTG), co is the control group (CTG only), cost is the average cost of resources used in the treatment and effect is the proportion of operative deliveries for non-reassuring fetal status averted by the addition of FPO to CTG monitoring (primary outcome). The difference between effect in the intervention group and that in the control group is the absolute risk reduction. Results of the ICER are presented in two complementary ways: numerically, in terms of a point estimate, with 95% confidence intervals, and diagrammatically, in terms of a cost-effectiveness plane, constructed by plotting the incremental costs and incremental effects of an intervention on a two-dimensional graph.20
Statistical uncertainty arises when a variable includes a range of estimates within which the true value is likely to be found. In economic analyses, the uncertainty surrounding the ICER is presented in a cost-effectiveness plane (Figure 1). This contains four quadrants representing four possible outcomes related to the costs and effects of the control and intervention groups. The first outcome, where an intervention is less costly but more effective than the control (CostsI < CostsC [CI > CC] and EffectsI > EffectsC [EI > EC]), is located in the southeast quadrant; the second outcome (CI > CC and EI < EC) in the northwest quadrant; the third outcome (CI > CC and EI > EC) in the northeast quadrant and the fourth outcome (CI < CC and EI < EC) in the southwest quadrant.
Figure 1. Cost-effectiveness plane. The point estimate of the ICER falls into the southeast quadrant. This represents the intervention (addition of FPO to conventional CTG monitoring) being more effective and less expensive than the control (CTG only).
Download figure to PowerPoint
To present the ICER on a cost-effectiveness plane, the point estimate of the change in costs is plotted on the y-axis, while the point estimate from the change in effects is plotted on the x-axis.
Sensitivity analysis, an important component of determining uncertainty,13 was also undertaken to examine the robustness of results when a key cost driver was varied. Throughout the trial, it was assumed that the FPO unit was used at full capacity. The sensitivity analysis explored the cost implications of varying the levels of capacity. Based on the 2402 fetal heart rate traces identified as being eligible for trial inclusion8 and the lengths of time the trial was conducted at each hospital, we estimated that there were approximately 282 non-reassuring fetal heart rate traces per hospital per year. It was therefore assumed that an FPO unit would be required for 5.4 births/week (per hospital), varied from four births to seven births/week for the purposes of the sensitivity analysis.
The location and spread of baseline demographic data were summarised using proportions, mean and SD or using median and interquartile range, as appropriate. Chi-square test/Fisher’s exact test, t tests or Mann–Whitney U test were used, as appropriate, to compare baseline patient characteristics for factors that may have influenced maternal or fetal morbidity. We planned to use regression modelling to address any differences in baseline patient characteristics, which may have influenced costings.21
As parametric assumptions do not necessarily hold for the ICER, bootstrap resamples were used to explore the statistical inferences of ICERs.20 Bootstrap resampling is based on repeated random sampling with replacement from an original sample to provide a collection of new pseudoreplicate samples, from which sampling variance can be estimated. In keeping with accepted practice, a total of 2000 replicates were taken from each sample and the bias-corrected accelerated method22 used to calculate 95% confidence intervals about the sample estimate of the ICER.
Statistical analyses were performed using SPSS for Windows, version 13 (SPSS Inc., Chicago, IL, USA) and @Risk software (Palisade Corporation, New York, NY, USA).