Introduction
An economic evaluation has been described as a comparative analysis of alternative courses of action in terms of both their costs and consequences.44 The economic evaluation for this study is detailed within the Health Economics Analysis Plan. In the URB trial, three economic evaluations were undertaken: a cost-effectiveness analysis (CEA), a cost–utility analysis (CUA) and a cost-consequence analysis (CCA). The economic evaluations were carried out from an NHS and patient perspective over the time horizon of one year.
A CEA equates an intervention’s extra cost to its extra effectiveness when measured using a natural or clinical outcome. For this CEA, the clinical outcome was the proportion of children who met the target difference in the CGI-I at 24 weeks – the primary outcome URB trial (see Chapter 2). The primary outcome of the CEA was incremental (i.e. the extra cost) per additional person meeting the target difference in the CGI-I at 24 weeks.
In addition to the CEA, a CUA was undertaken, in which the costs were compared to a measure that combines quantity and quality of life – in this instance quality-adjusted life-years (QALYs). The primary outcome of the CUA was the incremental cost per QALY at 52 weeks from the perspective of the caregiver and the child using the EQ-5D-5L health-related quality-of-life instrument.45 A secondary outcome measure for the CUA was the incremental cost per QALY gained at 52 weeks from the perspective of the child using the Child Health Unity 9D (CHU-9D) completed by a proxy quality-of-life instrument.46,47
The final form of economic evaluation undertaken was a CCA. Using this approach, the costs are compared to the outcomes from the trial using a balance sheet approach. The outcomes for the CCA are the primary and secondary outcomes that were collected as part of the URB trial.48 These measures are described in detail in Chapter 2 and their results are presented in Chapter 4. The purpose of the CCA is twofold. First, it provides additional context that can help understand what might drive the results seen in the CEA and the CUA. Second, both the CEA and CUA have a comparatively narrow consideration of what constitutes a benefit (or harm) of the intervention (e.g. the single measure of clinical effectiveness for the CEA and health-related quality of life for the CUA). Thus, both the CEA and CUA may not include some impacts that stakeholders may find important when making a decision about whether or not URB should be used in preference to LAA.
Methods
The economic evaluation was carried out using the NICE guidelines for health technology assessment that were applicable when the analysis of the URB data began.49 The version of this guidance has since been updated.50
Data collection
Intervention costs
The cost of the URB intervention was estimated from trial data and clinical expert opinion from the trial team. The URB group is delivered by two community-based professionals (facilitators) who have experience working with children with ASD and this was felt to be reflective of how the intervention would be delivered in routine practice. Professional time (the facilitators) required for this intervention was estimated using the clinical expertise of the trial team who designed the intervention. The amount of time and the type of staff required for group facilitations were estimated from expert opinion from the trial team who designed the intervention. Facilitator costs were derived by estimating the time required to prepare and run the group multiplied by the hourly rate for the facilitator based on published estimates.51 Printed materials costs were assumed based on trial teams’ estimates of printing rates.
For the LAA arm, the sessions were two hours long and were also delivered by a group facilitator trained and approved by the NAS. The NAS staff rates, the time required to run the group and the printed material costs were all provided by correspondence from the NAS.52
For both the URB and LAA arms, a per participant cost was derived by dividing the total intervention cost by the average number of participants per group. The number of participants per group was based upon an assumption of nine per group which was the most common group size in the Northumberland, Tyne and Wear NHS Foundation Trust which was the site that ran the most groups.
Each randomised participant was allocated a treatment cost based upon which arm they were randomised to. The intervention costs are summarised in Appendix 5, Table 30.
Service utilisation costs
To measure the services utilised by participants throughout the course of the trial, a Service Use Questionnaire (SUQ) was administered to participants. The SUQ asked participants how many times in the last 6 months they had accessed health services. The SUQ was administered at baseline, 24 weeks and 52 weeks follow-up. The SUQ included the number of hospital appointments, inpatient hospital stays (including reason and duration) and the type and number of healthcare appointments. Appointments differentiated between the different places where appointments were held, for example the participants’ home; by phone/virtually and within the practice. The SUQ also asked participants to record the medications that they were taking, the duration they had been taking them for and the dose of the medication.
Costs for hospital outpatient appointments, day cases and inpatient stays were obtained from published NHS reference costs.53 Medication costs were estimated from the British National Formulary (BNF);54 the access dates can be seen in Appendix 5, Table 37.
Costs for community-based appointments such as GP or distract nurse appointments were costed using the Unit Costs of Health and Social Care.51 Hospital inpatient unit costs were calculated by calculating the number of days that the participant had an inpatient stay by the unit cost derived from NHS reference costs.53 Day-case costs were multiplied by the number of days that the participant was in hospital for a day case. The unit costs used are detailed in Appendix 5, Tables 33, 34, 36 and 37. Medication costs were calculated by multiplying the unit costs of the correct dose of drug by the duration that the participant had been taking it (up to a maximum of 6 months to match the recall period of each administration of the SUQ).
A per participant cost for resource use was generated for each recall period at baseline, and for the 24 weeks and 52 weeks follow-up data collection points. This participant cost was inclusive of appointments, medications and hospital visits. Costs at 24 weeks and 52 weeks follow-up were summed so that an overall cost for each participant over the trial follow-up was estimated (note the costs at baseline were used as part of the ‘adjusted analysis’ as described below and described in Appendix 5, Table 35). Once costs for each participant were derived, then descriptive statistics such as a mean cost for each arm were generated along with suitable measures of variance, for example standard deviations.
Patient private care costs and time and travel costs
In the base-case analysis only NHS costs were included in the CEA and the CUA. However, to explore the impact of the costs borne by the participants in terms of privately paid for care and time and travel costs for accessing care were also included in a sensitivity analysis. This analysis assessed whether the inclusion of the patient costs affected the conclusions.
Participants were asked about any private costs they had incurred in managing their child’s symptoms. This including asking about services such as homeopathy, traditional Chinese medicine or private counselling costs. The cost of such private care was an assumption based on estimates from private providers to provide the service. These estimates can be seen in Appendix 5, Table 38.
In order to help estimate the time and travel costs of accessing care, a time and travel questionnaire (TTQ) was administered at baseline. In this questionnaire, the participant detailed the journeys made to appointments in the community and the hospital. For journeys that were undertaken by public transportation or taxis, participants were asked to provide details about any fares that were paid. Any journeys that were undertaken by car were costed using the fuel rate based on the business and self-employed expenses rate per mile from the Travel Analysis Guidance.55 In addition, any fees for car parking were also included in the journey costs. For hospital transport, costs were derived from the NHS reference costs 2019–20.53
In addition to travel costs, the costs for the participant’s time were included both for time to travel to and attend community-based and hospital appointments and for the management of days that the child had taken out of school. Participants were asked if they had to take time away from their work, care or studies to look after a child during school hours. An estimate was used for the number of days that would be taken off school based on a study by Totsika et al.56 This study assessed patterns of non-school attendance for those on the autism spectrum. They found a median absence of 2 days per 23 days. This was extrapolated over 6 months to give a total of 16 days per 6-month period. A school day was assumed to be 6 hours, and paid time was costed based on the Office for National Statistics (ONS) annual survey of hours and earnings (ASHE) average salary rate.57 Unpaid time was costed at the leisure time rate based on the study by Verbooy et al., estimate for hourly estimate value of leisure time and unpaid work.58 The time the participants spent attending the hospital and outpatient appointments was reported in the TTQ for community and hospital-based appointments and costed at the ONS rate.
The cost for time and travel for each type of service contact was calculated by taking the mean from the time and travel costs in the sample. This was then multiplied by the number of the appointments for each trial participant for both the hospital and the community. This was added to the private costs to make up the time, travel and private costs for each participant. An average cost for each trial arm was then calculated. The time and travel costs are summarised in Appendix 5, Table 31 and the school absence costs are summarised in Appendix 5, Table 32.
Estimation of effects
Clinical Global Impression – Improvement scale outcomes
The primary effectiveness outcome measure for the cost-effectiveness analysis is achieving at least the target difference in CGI-I at 24 weeks. The outcome is expressed as a percentage of children who achieved their targeted outcome. As noted above, the results for the CEA are expressed as the incremental cost per additional child reaching their target improvement in the CGI-I outcome.
Quality-adjusted life-year outcomes
The primary measure of effects for the CUA is the QALY derived from responses to the EQ-5D-5L (and CHU-9D in a secondary analysis). Participant related quality of life is measured at baseline, 24 and 52 weeks for both the EQ-5D-5L and CHU-9D. The EQ-5D-5L was completed by the child’s primary caregiver as a measure of their own health-related quality of life. The CHU-9D is by a proxy on behalf of the child as a measure of their health-related quality of life.
There is a new UK value set for the EQ-5D-5L which could be used to derive utility values.59 However, at present, this data set is still being validated and NICE advised mapping the values from the previous EQ-5D-3L data set, and as such, the results from the EQ-5D-5L were cross-walked to the EQ-5D-3L using the van Hout cross-walk,60 the recommended method at the time the analysis was started.
As noted above, the primary outcome for the CUA was the incremental cost per QALY based on responses to the EQ-5D-5L. The likelihood that this value is below a given value for society’s willingness to pay for a QALY (including NICEs threshold for a QALY gain of £20,000 and £30,000) was estimated.50 A secondary outcome for the CUA was the incremental cost per QALY based on responses to the CHU-9D. The likelihood that this value is below a given value from a range of society’s willingness-to-pay thresholds for a QALY was also estimated.
Balance sheet of outcomes
In addition to the CEA and the CUA economic evaluations, there was also to be a CCA, which lists the results of the primary and secondary trial outcomes alongside the costs for each arm. The list of the secondary outcomes in this trial are detailed in Table 10 in Chapter 4.
Analysis of costs and benefits
Continuous and count variables are expressed with appropriate descriptive statistics. This includes means and standard deviations and, for difference in outcome measures, confidence intervals. The relationship between the costs and the outcomes was analysed using a seemingly unrelated regression (SUR). A SUR is a type of regression analysis which permits the simultaneous estimation of costs data and effects data, calculated at an individual level. In addition, the SUR controlled for additional covariates (such as age and gender) that may affect costs, effects or both.61 The data were analysed in the Stata™ software program.62
The price year for costs and benefits was 2020 in Great British pounds (GBP). As the cost and benefits for this study were measured within 1 year, no discounting was applied aligning the NICE technology assessment guidelines.49
Missing data
For the complete case analysis, only those with complete case data were included. To be included in the complete case data set all relevant trial instruments had to be completed. This included both the visit and the service use questionnaire which had to be noted as complete on the visit record. In addition, to be considered completed the questionnaires must not have had large sections that were not completed (e.g. no values, positive or negative, were given in the entire section). For instruments used to derive (i.e. the EQ-5D-5L or CHU-9D) QALY values, data were missing if any one of the three time points of the baseline, 24 weeks or 52 weeks was not completed.
To account for those with missing data, multiple imputation was carried out. Costs and QALY were imputed using a chained imputation approach using paired mean matching approach using nearest neighbour. This approach was chosen because the data were missing at similar levels and costs and QALYs were considered to be interrelated variables.
In addition to the CEA and the CUA economic evaluations, there was also to be a CCA, which lists the results of the primary and secondary trial outcomes alongside the costs for each arm. The list of the secondary outcomes in this trial are detailed in Table 7.
Sensitivity analysis
Deterministic sensitivity analysis
Deterministic sensitivity analysis was carried out to assess the variability of different parameters on the outcomes of the economic evaluation. One of the key sensitivity analyses was to broaden the cost perspective to include the costs falling on participants and their families.
Stochastic sensitivity analysis
To assess the robustness of the study sampling, non-parametric bootstrapping was carried out to assess the uncertainty around the conclusions. Bootstrapping is a technique which resamples a single data set to create many simulated samples to assess statistical precision. In this case, this meant that the difference in net benefit was estimated for each simulated data set.
In this study, 1000 iterations of the bootstrapping procedure were performed. For the CUA, bootstrapping was carried out for the costs and both the QALY outcomes. The results of the bootstrapping URB were used to develop cost-effectiveness planes. For the CUA, the horizontal axis represents the difference in adult QALYs or child QALYs and the vertical axis represents the corresponding difference in costs.
For the CCA, appropriate measures of statistical variance, for example mean and standard deviation, were presented.
Coronavirus disease 2019
Due to the onset of the COVID-19 pandemic during this trial, the modality of the provision of the intervention changed from an in-person to an online format due to the constrictions around in-person meetings during the periods of lockdown. After a discussion with the trial team about how this would change the costs of the interventions, no important differences in costs could be identified for the different modes of delivery. The trial team used a free online platform, and the participants used their own electronic devices to participate. The clinical effectiveness analysis found that there was no evidence of a difference in the effectiveness in the interventions between modalities. As such, there was no specific sensitivity analysis for the impact of COVID-19 on the provision of interventions or its impact within the economic evaluation.
Reporting standard
The reporting of this economic evaluation was assessed against the consolidated health economic evaluation reporting standards (CHEERS checklist).63 This checklist notes the items to include when reporting economic evaluations of health interventions, and it was completed to make sure all relevant information was reported.
Results
Response rates
The response rates for the SUQ and the TTQ and are summarised in Table 14.
| Questionnaire | LAA group (n = 113) | URB group (n = 114) | ||||
|---|---|---|---|---|---|---|
| Data completeness (%) | ||||||
| Baseline | 24 weeks | 52 weeks | Baseline | 24 weeks | 52 weeks | |
| SUQ | 93% (n = 105) | 70% (n = 79) | 72% (n = 81) | 94% (n = 107) | 62% (n = 71) | 65% (n = 74) |
| TTQ | 79% (n = 89) | 69 % (n = 79) | ||||
As Table 14 shows, response rates for the SUQ and TTQ at baseline were high but did drop for the SUQ at subsequent follow-up time points. There appears to be no obvious differences in the pattern of missing data between trial arms The response rates for the EQ-5D-5L and the CHU-9D are summarised in Table 15. The results for the missing data for the EQ-5D-5L and the CHU-9D follow a similar pattern to the cost questionnaires (see Table 14). The results for data completeness for the primary outcome (the CGI-I) completeness data are also summarised in Table 15.
| Outcome | LAA group (n = 113) | URB group (n = 114) | ||||
|---|---|---|---|---|---|---|
| Data completeness (%) | ||||||
| Baseline | 24 weeks | 52 weeks | Baseline | 24 weeks | 52 weeks | |
| CHU-9D | 81% (n = 92) | 66% (n = 75) | 71% (n = 80) | 82% (n = 93) | 62% (n = 71) | 63% (n = 72) |
| EQ-5D-5L | 92% (n = 104) | 70% (n = 79) | 71% (n = 80) | 89% (n = 102) | 61% (n = 70) | 66% (n = 75) |
| CGI-I completeness of data % (n) | 71% (n = 81) | 65% (n = 74) | ||||
| Significant improvement in CGI-I score from baseline % (n =) | 12% (n = 10) | 15% (n = 11) | ||||
Total resource use
The summary of the use of health services captured by the SUQ are shown in Table 16.
| Cost category | LAA group | URB group | ||||
|---|---|---|---|---|---|---|
| Appointments | ||||||
| Mean (£) (SD) | Mean (£) (SD) | |||||
| 24 weeks | 52 weeks | Overall mean | 24 weeks | 52 weeks | Overall mean | |
| GP | 25 (37) | 20 (42) | 41 (64) | 36 (56) | 25 (46) | 61 (87) |
| GP out of hours | 6 (31) | 6 (34) | 14 (66) | 1 (9) | 2 (12) | 3 (16) |
| Community nurse | 2 (6) | 2 (10) | 3 (10) | 1 (4) | 11 (76) | 13 (80) |
| Community support worker | 1 (8) | 7 (44) | 9 (47) | 3 (22) | 2 (12) | 6 (26) |
| Social worker | 17 (55) | 18 (56) | 30 (101) | 13 (53) | 16 (60) | 31 (95) |
| CAMHS worker | 55 (191) | 316 (2310) | 406 (2571) | 115 (515) | 163 (529) | 301 (1043) |
| School nurse | 59 (285) | 139 (1080) | 221 (1333) | 62 (468) | 12 (34) | 17 (47) |
| Counsellor | 9 (66) | 2 (15) | 11 (82) | 4 (35) | 7 (46) | 12 (60) |
| NHS 24/NHS 111 phone line | 1 (3) | 1 (4) | 2 (4) | 1 (2) | 1 (4) | 2 (5) |
| Psychiatrist | 18 (127) | 18 (97) | 10 (60) | 76 (350) | 112 (405) | 208 (708) |
| Psychologist | 38 (102) | 50 (140) | 82 (178) | 100 (231) | 52 (161) | 161 (350) |
| Occupational therapist | 139 (471) | 121 (383) | 229 (578) | 268 (633) | 269 (852) | 531 (1356) |
| Paediatrician | 187 (291) | 195 (288) | 371 (440) | 243 (412) | 185 (399) | 444 (772) |
| Speech and language therapist | 259 (541) | 217 (502) | 448 (764) | 213 (418) | 408 (883) | 629 (1137) |
| Accident and emergency | 17 (48) | 27 (63) | 35 (85) | 6 (26) | 17 (72) | 26 (93) |
| Other consultations | ||||||
| Other | 83 (278) | 73 (198) | 166 (377) | 72 (227) | 28 (84) | 104 (256) |
| Medications | ||||||
| Atomoxetine | 2 (16) | 2 (17) | 2 (17) | 0 (0) | 0 (0) | 0 (0) |
| Mirtazapine | 0 (0) | 0 (0) | 0 (0) | 0 (0) | 0 (0) | 0 (0) |
| Propranolol | 0 (0) | 0 (0) | 0 (0) | 0 (0) | 0 (0) | 0 (0) |
| Sertraline | 0 (0) | 0 (0) | 0 (0) | 0 (0) | 0 (1) | 0 (2) |
| Fluoxetine | 0 (0) | 0 (0) | 0 (0) | 1 (1) | 0 (1) | 0 (2) |
| Other medications | ||||||
| Other | 83 (199) | 94 (202) | 146 (350) | 71 (186) | 88 (262) | 173 (410) |
A number of different cost elements are important drivers of total cost. This includes children and young people’s mental health services costs such as speech and language therapist costs and occupational therapy costs. Table 16 summarises the costs at 24 and 52 weeks; costs were also calculated at baseline from the preceding 6 months of the trial. As described in the method, these costs were used to adjust for baseline costs in the SUR and are summarised in Appendix 5.
Health-related quality of life
The utility values from both the EQ-5D-5L and the CHU-9D health-related quality-of-life instruments are summarised in Table 17.
| Outcome | LAA group (n = 113) | URB group (n = 114) | ||||||
|---|---|---|---|---|---|---|---|---|
| Mean (SD) | Mean (SD) | |||||||
| Baseline | 24 weeks | 52 weeks | QALYs | Baseline | 24 weeks | 52 weeks | QALYs | |
| CHU-9D | 0.80 (0.11) | 0.81 (0.11) | 0.84 (0.10) | 0.78 (0.11) | 0.80 (0.10) | 0.82 (0.11) | ||
| Child’s QALYs using CHU-9D (n = 121) | 0.82 (0.08) | – | 0.81 (0.08) | |||||
| EQ-5D-5L | 0.81 (0.19) | 0.82 (0.21) | 0.85 (0.17) | 0.73 (0.23) | 0.81 (0.21) | 0.79 (0.21) | ||
| Parents’ QALYs using EQ-5D-5L (n = 130) | 0.79 (0.18) | 0.82 (0.16) | ||||||
As Table 17 shows, there is a small amount of additional QALYs based on responses to the EQ-5D-5L in the URB arm compared with the LAA arm. The CHU-9D-based QALYs are slightly higher on average for the LAA arm compared with the URB arm. It should be noted that in neither case should these differences be over-interpreted as there was no evidence of any difference in QALYs for either measure of QALYs; this table does not show the results of any statistical tests for a difference.
Economic evaluation
Cost-effectiveness analysis
The results of the CEA analysis are shown in Table 18. As this table shows, URB is on average more costly and more effective than LAA. The incremental cost per additional child achieving the target difference of over £35,000 for URB compared with LAA. This deterministic analysis however does not reflect the imprecision in estimates of cost or effects.
| Data | Intervention | Cost (£)(CI) | ΔCost (£) | Children achieved target CGI-I | Δ Children achieved target CGI-I | ICER (ΔCost/ΔChildren achieved target) (£) CGI-I |
|---|---|---|---|---|---|---|
| Complete case data(n = 136) | LAA (n = 71) | £2655 (£1803 to £3507) | 0.13 | |||
| URB (n =65) | £3389 (£2359 to £4417) | £734 | 0.15 | 0.02 | £36,700 |
Complete case cost–utility analysis using quality-adjusted life-years based on responses to the EQ-5D-5L
The results of the complete CUA using QALYs based on responses to the EQ-5D-5L are shown in Table 19. The results are shown as a cost-effectiveness plane and cost-effectiveness acceptability curve (CEAC). These are graphic ways of demonstrating the probability of the intervention’s cost-effectiveness across a range of different values for society’s willingness to pay for an outcome. The cost-effectiveness plane is in the form of a scatter plot (Figure 4) and CEAC with the points expressed on an x and y axis (Figure 5).
| Data | Intervention | Unadjusted cost (£) (CIs) | Adjusted ΔCost (£) (CIs) | Unadjusted QALY(EQ-5D-5L)(CIs) | Adjusted Δ QALY(EQ-5D-5L)(CIs) | ICER (ΔCost/ΔQALY)(EQ-5D-5L) (£) | Probability URB cost-effective at different threshold values for society’s willingness to pay for a QALY | |||
|---|---|---|---|---|---|---|---|---|---|---|
| £0 | £20,000 | £30,000 | £50,000 | |||||||
| Complete case data(n = 129) | LAA (n = 68) | £2725 (1838 to 3611) | 0.79 (0.74 to 0.83) | 85% | 78% | 73% | 67% | |||
| URB (n = 61) | £3394 (2307 to 4480) | £685 (−525 to 1895) | 0.83 (0.79 to 0.87) | 0.001 (−0.04 to 0.043) | £685,000 | 15% | 22% | 27% | 33% | |
FIGURE 4 Cost-effectiveness plane for the complete data. Quality-adjusted life-years based on response to the EQ-5D-5L. (Figure not reproduced here; see the original article.)
FIGURE 5 Cost-effectiveness acceptability curve for the complete case data. Quality-adjusted life-years based on response to the EQ-5D-5L. (Figure not reproduced here; see the original article.)
The results reported in Table 19 for the complete case analysis show that the URB arm is on average more costly than the LAA arm. The CIs for the adjusted difference in costs have both positive and negative values, meaning that there is also a possibility that URB could be both more or indeed less costly compared with LAA. There is, on average, additional QALYs for URB compared with LAA. The CI however includes zero. Given the very small average increase in QALYs and more substantial increase in cost per patient on average for URB compared with LAA, the point estimate for the incremental cost per QALY gained for URB compared with LAA was almost £700,000. This estimate does not reflect the imprecision in estimates of costs of QALYs. To illustrate this, as described in the methods above, bootstrapping of incremental costs and QALYs was conducted, and these were used to plot how incremental costs and QALYs might vary (see Figure 4). As Figure 4 shows, for the majority of plots (85%) URB is more costly. The estimates of QALYs are evenly distributed (for 53% of the bootstraps URB provides more QALYs). This imprecision is reflected in the cost-effectiveness acceptability curve that is graphically shown in Figure 5 and in tabular form in the final four columns of Table 19. Figure 5 and Table 19 show that URB is unlikely to be considered cost-effective compared to LAA over all the threshold values for society’s willingness to pay for an additional QALY.
Secondary and sensitivity analyses
Table 20 and Figures 6 and 7 show the imputed sample for the EQ-5D-5L which is a broadly similar picture to the complete case analysis described above. With the larger sample available when data in imputed there is still a slightly larger additional effect on average but at a slightly lower additional average cost for the URB arm compared with the LAA arm. The incremental cost per QALY gained for this analysis is almost £45,000 and there is at best a fair chance that URB would be considered cost-effective over the range of threshold values for society’s willingness to pay for a QALY.
| Analysis | Data | Intervention | Unadjusted cost (£) (CIs) | Adjusted ΔCost (£) (CIs) | Unadjusted QALY(EQ-5D-5L)(CIs) | Adjusted ΔQALY(EQ-5D-5L)(CIs) | ICER (ΔCost/ΔQALY)(EQ-5D-5L) (£) | Probability URB is cost-effective at different threshold values for society’s willingness to pay for a QALY | |||
|---|---|---|---|---|---|---|---|---|---|---|---|
| £0 | £20,000 | £30,000 | £50,000 | ||||||||
| Cost–utility analysis of the imputed case data for the EQ-5D-5L | Imputed case data (n = 199) | LAA (n = 101) | £2644 (2028 to 3261) | 0.78 (0.75 to 0.82) | 84% | 63% | 56% | 46% | |||
| URB (n = 98) | £3005 (2287 to 3723) | £445 (−387 to 1278) | 0.84 (0.79 to 0.87) | 0.01 (−0.02 to 0.04) | £44,500 | 16% | 37% | 44% | 54% | ||
| Cost–utility analysis using the CHU-9D | Complete case data (n = 120) | LAA (n = 62) | £2394 (1701 to 3087) | 0.82 (0.80 to 0.84) | 97% | 93% | 90% | 85% | |||
| URB (n = 58) | £3250 (2205 to 4440) | £922 (−1567 to 2001) | 0.81 (0.79 to 0.83) | 0.001 (−0.002 to 0.023) | £922,000 | 3% | 7% | 10% | 15% | ||
| Cost–utility analysis of the imputed case data for the CHU-9D | Imputed case data (n = 181) | LAA (n = 90) | £2450 (1946 to 2954) | 0.82 (0.80 to 0.83) | −0.005 (−0.022 to 0.012) | LAA is dominant | 88% | 90% | 89% | 87% | |
| URB (n = 91) | £2822 (2068 to 3576) | £444 (−265 to 1256) | 0.80 (0.79 to 0.82) | 12% | 10% | 11% | 13% | ||||
| Cost–utility analysis of the imputed case data for the EQ-5D-5L including time, travel and private costs | Imputed case time, travel and private data (n = 199) | LAA (n = 101) | £2934 (2290 to 3577) | 0.78 (0.75 to 0.82) | 89% | 76% | 68% | 60% | |||
| URB (n = 98) | £3367 (2609 to 4126) | £526 (−339 to 1392) | 0.83 (0.80 to 0.86) | 0.006 (−0.027 to 0.039) | £87,667 | 11% | 24% | 32% | 40% | ||
FIGURE 6 Costs-effectiveness plane for the imputed case data for the EQ-5D-5L. (Figure not reproduced here; see the original article.)
FIGURE 7 Cost-effectiveness acceptability curve for the imputed case data for the EQ-5D-5L. (Figure not reproduced here; see the original article.)
Cost–utility analysis using the Child Health Unity 9D and complete case analysis
The results for the complete case CUA using the CHU-9D values for the QALYs are summarised in Table 20 and Figures 8 and 9. The results show a very small effect in favour of URB with a larger cost differential compared with the EQ-5D-5L analysis reported above. Overall, the pattern of findings is the same and the URB intervention is unlikely to be considered cost-effective over a range of different willingness to pay for a QALY.
FIGURE 8 Costs-effectiveness plane for the complete case data for the CHU-9D. (Figure not reproduced here; see the original article.)
FIGURE 9 Cost-effectiveness acceptability curve for the complete case data for the CHU-9D. (Figure not reproduced here; see the original article.)
Cost–utility analysis using the Child Health Unity 9D and imputed data
The results for the imputed CHU-9D sample are shown in Table 20, Figures 10 and 11. These data are broadly similar to those already presented. Although LAA is slightly more effective on average and has a lower cost on average, the likelihood of URB being cost-effective compared with LAA over the range of threshold values for society’s willingness to pay is low.
FIGURE 10 Cost-effectiveness plane for the imputed case data for the CHU-9D. (Figure not reproduced here; see the original article.)
FIGURE 11 Cost-effectiveness acceptability curve for the imputed case data for the CHU-9D. (Figure not reproduced here; see the original article.)
Cost–utility analysis for the imputed case data for the EQ-5D-5L where patient costs are included
The results for this analysis are shown in in Table 20, Figures 12 and 13. The inclusion of the cost borne by the child and carers does not change the pattern of results already seen. Compared with LAA, URB is unlikely to be considered cost-effective over the range of threshold values for society’s willingness to pay for a QALY.
FIGURE 12 Costs-effectiveness plane for the imputed case data for the EQ-5D-5L including time, travel and private costs. (Figure not reproduced here; see the original article.)
FIGURE 13 Cost-effectiveness acceptability curve for the imputed case data for the EQ-5D-5L including time, travel and private costs. (Figure not reproduced here; see the original article.)
Cost-consequence analysis
In addition to the CEA and CUA, a CCA comparing the costs to the secondary outcomes was carried out. The results can be seen in Table 21.
| Impacts that favour LAA | Impacts that favour URB | ||
|---|---|---|---|
| Measure | Value | Measure | Value |
| RBQ Teacher Mean Total score. URB -LAA change from baseline at 52 weeks (95% CI) | 0.18 (0.05 to 0.31) | – | – |
| No evidence of a difference | |||
| Lower adjusted costs on average (95% CI) | £445 (−387 to 1278) | ||
| RBQ-2 Sensory Motor Behaviour. URB-LAA change from baseline diff (95% CI) | −0.01 (−0.09 to 0.07) | VABS-3 Socialisation. URB-LAA change from baseline at 24 weeks (95% CI) | 0.21 (−1.94 to 2.37) |
| RBQ-2 Insistence on Sameness. URB -LAA change from baseline at 52 weeks (95% CI) | 0.02 (−0.08 to 0.12) | VABS-3 ABC. URB-LAA change from baseline at 24 weeks (95% CI) | 0.2 (−1.28 to 1.68) |
| RBQ-2 Mean Total score. URB-LAA change from baseline at 52 weeks (95% CI) | 0.01 (−0.06 to 0.09) | PSE. URB-LAA change from baseline at 52 weeks (95% CI) | 0.19 (−0.07 to 0.44) |
| VABS-3 Communication. URB -LAA change from baseline at 24 weeks (95% CI) | −0.94 (−3.26 to 1.4) | APSI Total. URB-LAA change from baseline at 52 weeks (95% CI) | −0.81 (−3.2 to 1.59) |
| VABS-3 Daily Living Skills. URB -LAA change from baseline at 24 weeks (95% CI) | −0.32 (−2.32 to 1.65) | AFEQ Experience being parent. URB-LAA change from baseline at 52 weeks (95% CI) | 0.17 (−1.5 to 1.85) |
| WEMWBS. URB-LAA change from baseline at 52 weeks (95% CI) | −1.3 (−3.45 to 0.86) | AFEQ Family life. URB -LAA change from baseline at 52 weeks (95% CI) | −0.04 (−1.39 to 1.32) |
| AFEQ Total. URB-LAA change from baseline at 52 weeks (95% CI) | 0.34 (−3.45 to 4.12) | ||
| AFEQ Child development understanding social relationships. URB-LAA change from baseline at 52 weeks (95% CI) | 0.17 (−1.5 to 1.85) | ||
| AFEQ Child symptoms. URB-LAA change from baseline at 52 weeks (95% CI) | 0.74 (−0.39 to 1.86) | ||
In addition to the CEA and CUA, a CCA comparing the costs to the secondary outcomes was carried out. The results can be seen in Table 21. The value representing the impact of each of the measures is derived from Table 10 in the previous chapter. For the majority of outcomes there is no evidence of a difference, although for the majority of these the CI are wide enough to include clinically or economically important differences. The only measure where there was evidence of a difference was the RBQ Teacher Mean Total score, which favours the LAA intervention.