Economic evaluation findings

Sample characteristics

Table 23 shows the responses to questionnaire data on individual health at baseline and 24-month follow-up for both trial arms. The majority of participants in both trial arms reported having a long-term illness, health problem or disability. At both baseline and 24-month follow-up, and for both trial arms, the largest proportion in each arm described their health status over the previous 6 months as good. The proportion in each arm reporting a long-term illness, health problem or disability, and a health problem limiting their daily activities, decreased over the 24 months of the trial. However, comparing the average baseline EQ-5D-3L score for those participants with missing EQ-5D-3L at 24-month follow-up with the average score for the complete study population showed that those with missing follow-up data had reported a substantially lower EQ-5D-3L at baseline [intervention group 0.589 (complete study population) vs. 0.529 (participants with missing 24-month follow-up EQ-5D-3L score); control group 0.583 vs. 0.457].

Cost–consequences analysis

Costs

Table 24 shows the durations of, and associated costs for, the WRAs’ activities required to deliver the intervention to participants in the intervention group. These activities included the home visit itself as well as additional telephone calls, letter or e-mail writing, administrative tasks (such as form filling or referring to other agencies), and other or unspecified activities, as well as travelling to the participants’ homes. Some WRAs recorded durations of 0 minutes for activities or distances of 0 miles for travel. These were considered valid responses as, on occasions, WRAs reported that they combined appointments when these were in close geographical proximity on the same day for efficiency.

The average total cost per participant for delivery of welfare rights advice was calculated as £43.76; 38% of these costs (£16.80) were travel costs associated with WRAs travelling to participants’ homes.

Financial benefits

Twelve financial benefits and allowances recorded by WRAs in the CCS were found to have been awarded to participants in the intervention group. Means-tested benefits included Council Tax Benefit, Housing Benefit, Pension Credit (guarantee, savings) and Employment Support Allowance, while non-means tested benefits were Disability Living Allowance (care and mobility components), Attendance Allowance (at high and low rate), Carer’s Allowance, Industrial Injuries Disablement Benefit and Severe Disablement Allowance. For all except two benefits (Employment Support Allowance and Severe Disablement Allowance), information on durations were available from the CCS for participants in the intervention group. The 10 financial benefits for which there was information on the duration of award (i.e. excluding Employment Support Allowance and Severe Disablement Allowance) were therefore considered when comparing the effectiveness of the intervention between trial arms.

Given the limited information collected in the CCS, partner benefits, one-off payments for financial benefits and pending claims were not considered in the initial analyses. In the CCS data, 10 financial benefits received by partners were recorded, but there were no data available to estimate the value of these over the 24-month follow-up period. For instance, for most of the recorded benefits, the only information available was on the recipient of the benefit. No information on the date of award or the weekly amount was recorded. No information on one-off payments was recorded in the CCS. The analysis also excluded pending claims, as the impact of these would not be observed in the 24-month follow-up period.

Table 25 reports the number of financial benefits newly awarded during the 24-month follow-up period and the average total amount (as well as the median and IQR) gained per participant in each trial arm when there were complete data available (a complete-case analysis), as well as the difference in average total amount. The benefits with the highest number of new claims in the intervention group were Council Tax and Attendance Allowance paid at the higher rate, while Council Tax Benefit, Housing Benefit and Disability Living Allowance (care component) were the most frequent new benefits in the control group. There was evidence of statistically significant differences between the trial arms in the weekly amounts for Disability Living Allowance (mobility component) and Carer’s Allowance, for which participants in the control group reported receiving significantly higher weekly amounts.

Contrary to our expectations, the total average value of financial benefits was higher for control group participants awarded new financial benefits than for those in the intervention group. The difference between the intervention and control arms in the total mean amount for newly awarded financial benefits per participant was –£451 (95% CI –£1892 to £991). Although there was no evidence of a statistically significant difference between groups, the CIs were wide enough to include economically important differences favouring either group.

These findings did not change when the discount rate was increased to 3.5% (Table 26).109 Sensitivity analyses capturing extreme case scenarios assessed the confidence around the imputed missing information on durations in the control group by imputing the missing value with values of either the first (Table 27) or the third quartile (Table 28) from the distribution of duration since receipt of benefit in the intervention group. In both cases, usual care was estimated to be more effective with regard to the average total amount gained per participant across all financial benefits, although this was not statistically significant.

Missing data were observed for some participants in the intervention group because of missing ‘date awarded’ information. These data were not replaced because no information was available on whether or not the initial claim was successful (i.e. whether or not the benefit was eventually awarded). The average total amount gained per participant and per benefit was estimated by combining information on newly awarded benefits, weekly amounts obtained from questionnaire data for both trial arms and median durations observed in the intervention group.

Non-financial benefits

Non-financial benefits recorded by WRAs in the CCS included the following items in addition to some aids and adaptations: Blue Badge (disabled driver), Warm Zones referral (including amount) and help with health-care costs (optician prescription charges, dental treatment charges, NHS travel costs) (Table 29).

As Table 29 shows, the numbers of new non-financial benefits recorded on the CCS by WRAs for participants in the intervention group were substantially different from the numbers of financial benefits reported by participants in the questionnaire.

To avoid an underestimation of effects in the intervention group, the analysis of non-financial benefits considered the same data source (i.e. questionnaire) for both the intervention and the control group, and so faces the same potential recall bias, thus providing a fair comparison between the two groups.

There was no evidence of statistically significant differences in the number of newly awarded benefits between trial arms, except that significantly more participants in the control group received general help with insulation costs.

One element of non-financial benefits that participants could receive was new aids and adaptations that helped them to remain independent at home. A full list of all items can be found in Table 30. This information was taken from the responses to the questionnaire in order to provide a fair (less biased) comparison. There was no evidence of any statistically significant differences in the number of newly awarded items between trial arms, with the exception of a ‘special telephone’, which was more frequently received in the intervention group.

Table 31 reports the average total value for aids and adaptations per participant. On average, the intervention group received £134 per participant more than the control group in the value of non-financial benefits in the form of aids and adaptations over the 2-year follow-up period (95% CI –£582 to £850).

Health-related quality of life

Changes in HRQoL were captured by a comparison of EQ-5D-3L scores between trial arms. EQ-5D-3L responses are broken down into the responses for each of the five questions so that it is possible to observe how responses to each question may differ between trial arms (see Table 37, Appendix 20). The observed pattern in responses to each of the questions at both baseline and 24 month follow-up was similar between trial arms.

Table 32 reports the mean EQ-5D-3L scores by each group. When converting the responses to the EQ-5D-3L into a score, the 36 participants who died during the trial follow-up period were scored as 0 at the 24 month follow-up, and these data have contributed to the estimation of mean scores. The number of deaths was evenly distributed between the randomised groups.

Not all participants who were alive at the end of trial follow-up completed the EQ-5D-3L. To understand if there might be some informative reason why there was not a response to the EQ-5D-3L, the correlation between whether or not EQ-5D-3L was missing and the primary study outcome, CASP-19, was estimated. This analysis showed that there was no evidence that missing EQ-5D-3L values were highly correlated with the CASP-19 (Pearson’s correlation coefficient –0.0886; p = 0.0668). Therefore, it was assumed that missing EQ-5D-3L responses for those who were not known to have died were missing at random.

For both trial arms, EQ-5D-3L scores were slightly higher at 24 months. QALYs calculated from EQ-5D-3L data at both baseline and 24 months were slightly higher on average in the intervention group after adjusting for differences in baseline EQ-5D-3L, age and gender (see Table 32), but there was no evidence of a statistically significant difference between trial arms. The CI is wide enough, however, to include clinically important differences favouring either trial arm.

Balance sheet

The results of the cost–consequence analysis are presented in the balance sheet (Table 33), indicating which arm of the intervention they favour (i.e. in which trial arm a significant positive difference in a particular benefit was observed).

Cost–utility analysis

On average, the (adjusted) intervention was found to be more costly and more effective than usual care (Table 34).

However, the differences in QALYs gained were not significant (p = 0.966). The point estimate of incremental cost per QALY gained was £1914. Nevertheless, there is considerable statistical imprecision surrounding these data, and bootstrapping was used to estimate the precision surrounding estimates of cost, QALYs and incremental cost per QALY. The incremental cost-effectiveness plane (Figure 6) visually plots the differences in mean costs and effects from each of the 1000 bootstrap iterations and indicates the level of uncertainty around cost and QALY point estimates (as previously reported by the 95% bootstrapped CIs in Table 34). Every point above the x-axis indicates that the intervention is more costly than usual care. Every point left of the y-axis indicates that the intervention is less effective than usual care, and right of the y-axis is more effective.

FIGURE 6 Cost-effectiveness plane. (Figure not reproduced here; see the original article.)

The statistical imprecision surrounding estimates of cost-effectiveness are presented in terms of a CEAC (Figure 7). A CEAC shows the one-sided probability that the intervention is cost-effective at any given value for society’s WTP for a QALY gained. This is analogous to the upper interval in a CI; the lower interval is irrelevant if we believe that the intervention is cost-effective at the upper interval because it must therefore be cost-effective at the lower interval.

FIGURE 7 Cost-effectiveness acceptability curve. (Figure not reproduced here; see the original article.)

In the CEAC we calculate the number of bootstrap iterations where the intervention arm is likely to be considered cost-effective compared with control over a range of values for society’s WTP for a QALY gain. As we have 1000 bootstrap iterations, we can estimate the probability that the intervention is cost-effective over the range of values. Furthermore, if the intervention is not cost-effective compared with the control then the control is cost-effective.

The probability that the intervention would be cost-effective, should society be willing to pay £20,000 per QALY gained, was around 63%. This suggests, given current evidence, that in economic terms and taking a NHS and local authority perspective we are almost indifferent to whether or not the intervention is implemented.

From an individual perspective, however, the value of the benefits can be considered a gain, albeit one that has not clearly translated into improved health over the 2-year time horizon of the study.

These results were robust to changes in the discount rate and higher costs associated with the delivery of the intervention.

The base-case analysis was based on complete cases. When missing EQ-5D-3L data were imputed using multiple imputation, the QALYs gained in the control group were, on average, 0.019 higher (95% CI –0.068 to 0.029). Therefore, the intervention was, on average, more costly and less effective than the control, and hence the intervention was dominated by usual care (Table 35).