Methods

Although the cost of GTN is low, there may be further costs associated with its administration and the monitoring and management of women thereafter. It is therefore important in the context of scarce maternity resources to explicitly quantify the net costs (or cost savings) associated with its use. Provision was therefore made in the GOT-IT protocol to carry out a simple cost analysis using clinical and resource-use data collected for individual participants recruited to the trial. This analysis explicitly quantifies the difference in mean costs between the active intervention and placebo groups.

Resource use associated with the alternative management strategies was estimated from the time of randomisation through to 6 weeks post childbirth. This included:

1. staff time for administration of the study drug and monitoring of the patient until time of placenta delivery or transfer to theatre for manual removal of placenta
2. resource use associated with complications arising following administration of the study drug or placebo
3. subsequent costs associated with delivery of the placenta (either spontaneously or operatively)
4. costs associated with the postnatal stay (to discharge)
5. subsequent health service contacts potentially relating to retained products of conception up to 6 weeks post discharge.

The time from administration of the study drug to spontaneous delivery of the placenta (or the decision to proceed with manual removal), and the incidence of complications following administration of the drug, were collected using the trial case report forms. Information about health service use in the 6 weeks following discharge was collected via a 6-week postnatal record check and several health service resource-use questions were included in the participant postnatal questionnaire (also at 6 weeks). Resource use was valued using routine sources of nationally relevant unit costs.^40–44

Mean costs are summarised by treatment allocation group following the principles of intention to treat, and the incremental cost (cost saving) associated with the use of GTN is estimated using linear regression with cluster robust CIs to adjust for potential centre effects. The cost data is presented alongside the primary and secondary outcome data in a cost–consequence balance sheet, indicating which strategy each outcome favours.

Measurement and valuation of resource-use data

The time from randomisation to delivery of placenta or transfer to theatre for manual removal of placenta was derived from the trial case report forms and costed using the unit cost per patient contact hour for someone on band 6.⁴⁰ This multiplier includes an overhead and capital apportionment. For those not requiring manual removal of placenta or transfer to theatre for other reasons, the remaining maternal stay was costed using a national average bed-day cost following vaginal deliveries, applied to actual duration of stay following delivery of the placenta. This unit cost was taken as relevant excess bed-day cost obtained from the NHS Reference Costs 2015 to 2016.⁴¹ Because we did not have detailed data on mode of vaginal delivery for individual patients, we applied a weighted average (by national activity levels) of the excess bed-day costs for all health-care resource-use groups relating to normal and assisted deliveries: NZ30-34 and NZ40-44. Drug costs for GTN were also applied to those randomised to the GTN group (Table 14).⁴²

TABLE 14

Identification, measurement and valuation of resource use: intervention

Tables 15–17 summarise the data sources used for the costing of hospital resources (see Table 15), primary care (see Table 16) and secondary care (see Table 17). For women who were transferred to the theatre but did not subsequently require a manual removal of placenta, the cost of theatre resources was estimated by applying a running cost per hour of obstetric theatre time obtained from Scottish Health Service Costs (2016) data.⁴³ For women requiring a blood transfusion, a unit cost per red cell unit transfused was applied, assuming an average of two units per woman.⁴⁴ For women who were transferred to theatre and received manual removal of placenta, these episodes were costed using the NHS reference cost for such episodes. Manual removal of the placenta maps to health-care resource-use group code NZ27Z (postnatal therapeutic procedures).⁴¹ The unit cost for this health-care resource-use group code reflects the average cost of these procedures across the NHS. The time and date of discharge were additionally used to estimate the overall health-care resource-use group code-based reference cost for each manual removal of placenta episode.

TABLE 15

Identification, measurement and valuation of resource use: hospital resource

TABLE 17

Identification, measurement and valuation of resource use: secondary care

TABLE 16

Identification, measurement and valuation of resource use: primary care

The 6-week postnatal record check provided data on any readmissions to hospital deemed to be potentially relevant to retained placenta or study drug. All readmissions were costed using appropriate health-care resource-use group codes combined with length of stay. Finally, the postnatal patient questionnaire provided information on participants’ use of community midwifery, health visitor and primary care services up to 6 weeks post discharge. Reported use of these services was costed using national unit cost data.⁴⁰ In addition, the questionnaire collected data on any further outpatient appointments. Outpatient attendances were costed using the average unit price for outpatient appointments obtained from national sources.⁴⁰ The questionnaire was piloted with 15 women as part of the qualitative study at Edinburgh Royal Infirmary prior to finalisation (see Chapter 3). On the basis of these interviews, the wording of the questionnaire was refined slightly to avoid capturing routine primary care and health visitor contacts relating to the health of the baby. Tables 14–17 summarise the type of resource use captured and the associated sources of measurement and valuation. All costs were expressed for the 2015–16 financial year.

Analysis of cost data

Total costs were estimated for each woman enrolled in the trial as the sum of each cost element. These are summarised by treatment allocation group (by intention to treat) and broken down into the following categories: intervention and monitoring costs [until delivery of the placenta (non-manual removal of placenta) or transfer to theatre], theatre costs for non-manual removal of placenta procedures, manual removal of placenta costs, total episode costs (randomisation to discharge) and postdischarge costs (primary and secondary care). The mean differences in cost between groups was estimated using ordinary least squares regression. Cluster robust standard errors and CIs were implemented to account for possible within-centre correlation. All analyses were conducted in Stata 13^® (StataCorp LP, College Station, TX, USA).

Further sensitivity analysis was conducted to assess the robustness of the findings to the costing approach. For the first sensitivity analysis [sensitivity analysis 1 (SA1)], we costed all manual removals using the running cost per hour of obstetric theatres in Scotland⁴³ and costed the postnatal admission using the average bed-day cost following vaginal deliveries. For the second sensitivity analysis [sensitivity analysis 2 (SA2)], we costed manual removals using the difference between the NHS reference costs for vaginal deliveries with (NZ32, NZ33, NZ42 and NZ43) and without (NZ30, NZ31, NZ40 and NZ41) postnatal surgery.⁴¹

The number of missing data was very low for major cost drivers associated with the initial delivery episode and readmissions to hospital, but substantial for postdischarge outpatient and primary care costs up to 6 weeks. Alternative assumptions about mechanisms for missing data were applied depending on the variable; missing elements of primary care use, where participants had responded to some resource-use questions but not others, were assumed missing not at random but because of irrelevance to the participant. In this case, zeros were assigned. Other resource-use variables were assumed missing at random. We originally planned to use multiple imputation to generate multiple plausible values for these missing data items, but because data was complete for over ≈90% of the major cost drivers, this was deemed unnecessary. The postdischarge outpatient and primary care costs up to 6 weeks formed only a very small component of the overall costs.

Differences in costs are presented alongside the statistical comparisons of primary and secondary efficacy and safety outcomes in the form of a cost–consequence balance sheet. This provides a summary of the strategy favoured on cost and all the other outcomes of interest. All effects on the clinical, patient-oriented and safety outcomes were derived as detailed in the statistical analysis section (see Chapter 5).

Results

Table 18 summarises the resource use associated with the index hospital episode, from time of randomisation to time of discharge. There were no obvious notable differences in elements of resource use associated with the hospital stay. Manual removal of placentae were slightly more frequent in the GTN group, as were blood transfusions.

TABLE 18

Resource use associated with the delivery episode by treatment allocation group

Table 19 summarises further secondary care resource use between discharge and 6 weeks post discharge. Again, there are no major notable differences between the groups.

TABLE 19

Secondary care resource use to 6 weeks post discharge

Table 20 summarises primary care usage to 6 weeks post discharge. Again, there were no notable differences in elements of resource use by treatment allocation. Outpatient visits were slightly more common in the GTN group than in the placebo group, and readmissions were slightly more common in the placebo group than in the GTN group, but the length of stay among those readmitted was higher in the GTN group.

TABLE 20

Primary care resource use to 6 weeks post discharge

The costs associated with the initial delivery episode, from time of randomisation to time of discharge, are summarised in Table 21 by treatment allocation. Overall, the costs were similar between groups, with the total episode cost being slightly higher in the GTN group than in the placebo group. There were no obvious differences between the groups in terms of the postdischarge hospital costs (Table 22) or postdischarge primary care costs (Table 23). The estimates of postdischarge primary care costs and outpatient costs are subject to a substantial amount of missing data, because of reliance on participant responses to the 6-week postnatal questionnaire.

TABLE 21

Initial hospital episode costs by treatment allocation

TABLE 22

Postdischarge hospital costs by treatment allocation

TABLE 23

Postdischarge primary care costs by treatment allocation

Incremental results

Table 24 summarises total costs by category and treatment allocation, and the estimated between-group difference. It can be noted that there are no significant between-group differences in any of the major cost categories. Directionally, there is a tendency towards slightly higher hospital episode costs in the GTN group than in the placebo group and this difference is consistent across both the approaches used to costing (SA1 and SA2).

TABLE 24

Difference in NHS costs by category and treatment allocation

Table 25 presents the main cost outcomes alongside the principal clinical, patient-oriented and safety outcomes, and secondary outcomes found to differ significantly between groups. Although the costs are directionally slightly higher in the GTN group than in the placebo group, there are no significant differences in the primary clinical, patient-oriented or safety outcomes. Furthermore, there is some evidence of a detrimental effect of GTN on several secondary outcomes, including the need for blood transfusions, side effects and blood pressure or heart rate.

TABLE 25

Cost–consequences balance sheet

Summary/discussion

The economic analysis is based on resource-use data that were collected for individual participants who were enrolled in the trial, combined with nationally relevant unit cost data. The strengths of the analysis include the secure randomisation, low loss to follow-up on the elements of resource use that are the most significant cost drivers and the generalisability of the unit costs applied. A potential limitation of the analysis is the reliance on reference costs as opposed to detailed bottom-up costing, because reference costs can lack the precision to capture the cost impact of finer differences in patterns of resource use between groups. To overcome this limitation we have presented more detailed data on resource use separately from costs, and these data show very similar levels and patterns between the treatment allocation groups. We also conducted further sensitivity analysis to assess the impact of utilising different methods to cost the delivery episode. A further limitation of the trial was the low level of follow-up on postdischarge resource use of primary care and hospital outpatient care. However, it can be noted from the available data that these costs were a relatively minor component of the overall costs and did not differ significantly between groups. In addition to the above, the economic analysis lacked an appropriate multidimensional measure of value by which to compare the alternatives. This is primarily attributable to the limitation of quality-adjusted life-years in this clinical context. Therefore, the analysis relied on a cost–consequence approach, which compared the different costs and multiple consequences of the alternative interventions without generating an incremental cost-effectiveness ratio or net benefit statistic.

However, considering the estimated costs in the context of the clinical effectiveness and safety findings, the data provide a clear message. The use of GTN for the treatment of retained placenta is not cost-effective compared with standard practice. Although there is a non-significant trend towards slightly higher costs in the GTN group, there are no significant differences in the primary clinical, patient-oriented or safety outcomes. Furthermore, directionally these primary outcomes favour placebo, and several secondary outcomes also point towards an increased side-effect profile for GTN over placebo, a possible safety concern with respect to blood pressure or heart rate and an increased number of blood transfusions required.