Outcome 2 (health coaching randomised trial and cost-effectiveness analysis)

Recruitment and retention

Patient flow is outlined in Figure 8. In total, 1306 patients met the eligibility criteria and 504 were selected for health coaching. Rates of follow-up for the various outcome measures are shown in Figure 8. There was evidence of an imbalance in attrition, with slightly higher rates of loss among patients selected for the intervention.

FIGURE 8 The PROTECTS trial CONSORT diagram. QoL, quality of life; SD, standard deviation. Reproduced from Panagioti et al.147 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. (Figure not reproduced here; see the original article.)

Table 34 shows the table of baseline characteristics comparing patients selected for the health coaching with those eligible but not selected, as a check on the success of the randomisation.

Consent to participate in health coaching

Signed consent to the offer of health coaching among those eligible was initially 41% (207/504), although only 189 actually received calls (38%). Logistic regression exploring baseline factors associated with consent to health coaching is presented in Table 35. In multivariate analyses, only younger age and higher education predicted consent to take up the health coaching intervention.

The process of health coaching in patients who consented

The defined ‘dose’ of the PROTECTS intervention was monthly calls of around 20 minutes for a period of 6 months, with the option for additional calls to deal with complex patients or issues of risk.

A total of 189 out of 504 (38%) people who were randomised to the health coaching intervention agreed to take part and were referred to the health coaching team from July 2015 to March 2016. In terms of adherence, 167 out of 189 (88%) participants who consented to the intervention received four or more telephone calls.

Results on primary and secondary outcomes at follow-up

Table 36 shows the patient-reported outcomes for patients selected for the intervention and those not selected. There were no significant differences on any primary or secondary outcome.

In the CACE analyses, effect sizes were higher, but none of the outcomes showed significant benefits to health coaching (see Table 36).

Sensitivity analyses provided no evidence for the results being substantively influenced by model assumptions (Table 37).

Economic results

Complete-case analysis requires data on age and sex at baseline, and EQ-5D-5L responses and resource utilisation at all four time points. Complete data necessary for the economic analysis were available for 45% of the sample (584/1306).

Table 38 presents the baseline characteristics of the full sample compared with the complete-case sample.

Resource utilisation and costs: health coaching

The resources required to deliver health coaching are presented in Table 39. They consisted of training sessions on intervention delivery, staff supervision, information materials for staff and patients, and the cost of coaching sessions.

The main training session was delivered by project staff and lasted 2 full days. A second half-day training session focusing on long-term conditions was delivered by a GP. There were five additional training sessions, again delivered by project staff, each lasting 90 minutes. Thirteen 1-hour supervision sessions were conducted with project staff. Manuals and scripts for health coaches and information sheets and notes pages were printed for participants. Telephone coaching calls lasted between 15 and 25 minutes, with an additional 10–15 minutes of staff preparation before or after the call. On average, each participant had one unanswered or rearranged call, which utilised an additional 10 minutes of staff time. The average cost per individual receiving the full course of health coaching (six calls) was £148.27, of which £44.38 related to training, supervision and paperwork and £103.89 related to delivery of health coaching.

Additional resource utilisation over trial follow-up

In addition to the direct costs of the health coaching, the economic analysis also considered wider NHS resource utilisation. Table 40 reports the average utilisation by resource category for the complete-case sample, separated by treatment arm, and summarised as the mean number of contacts with each service per individual.

Although there was variation in the use of services over time and between treatment arms, some consistent patterns can be seen. The most frequently utilised category across both treatment groups was outpatient appointments, followed by GP appointments.

The average number of emergency admissions (short stays) and A&E attendances was lower for the health coaching arm at all three time points, with the differences between the treatment arms increasing over time. The health coaching arm also had lower emergency admissions (long stays) and day cases at all three time points. Elective admissions and outpatient appointments were higher at all three time points among the health coaching arm, with the difference in outpatient appointments increasing over time.

Resource use was higher at 18 months than it had been at 12 months for all resource use categories in both treatment arms. However, in the health coaching arm, emergency admissions, day cases, outpatient appointments and A&E attendances fell at 12 months compared with 6 months, before rising again. Although emergency admissions (short stays), day cases and outpatient appointments also fell in the usual care arm, no such drop was observed for emergency admissions (long stays) or A&E attendances, with use rising at each consecutive time point in the usual care arm.

Overall, there was a pattern of greater use of emergency care among the usual care arm, whereas health coaching patients utilised more planned services.

To cost both arms, unit costs were applied to individual utilisation. Table 41 presents a list of all unit costs used, together with their source. Table 42 combines the resource utilisation of the complete-case sample with the unit costs to obtain the average costs associated with this resource utilisation over the trial period.

The most costly category of resource use was outpatient appointments, followed by elective admissions and GP appointments. These are all planned care services, the costs of which were higher in the health coaching group. Conversely, the costs of emergency admissions (short and long stays), day cases and A&E attendances were higher in the usual care group. Overall, mean costs were higher in the health coaching group (£4000.88) than in the usual care group (£3424.16).

Intervention costs in the health coaching averaged £79.29. This was lower than the £148.27 estimated for full treatment because not all individuals took up the offer or completed the full course.

Outcomes

Table 43 summarises the EQ-5D-5L utility scores at each time point and the total QALYs gained over the 18-month follow-up period for the complete-case sample. On average, patients who received health coaching reported slightly lower health-related quality of life at baseline than those who received usual care (mean utility 0.696 and 0.708, respectively). This value fell steadily over the follow-up period in the usual care group, reaching 0.664 at the 18-month follow-up, but in the health coaching group remained relatively unchanged at the end of follow-up, at 0.691. The mean unadjusted QALYs for usual care and health coaching over the study period were 1.105 and 1.124, respectively.

Cost-effectiveness analysis: full sample with imputation

Table 44 presents the adjusted estimates of the effect of the offer of health coaching on the incremental costs and QALYs over and above usual care in the full sample with imputed data. This analysis controls for age, sex and baseline utility. Health coaching is associated with a mean incremental total cost increase of £150.583 (95% CI –£470.611 to £711.776) and a mean incremental QALY gain of 0.019 (95% CI –0.006 to 0.043).

Although there are no statistically significant differences in either costs or QALYs, the point estimate of the ICER is £8049.96 per QALY. This would represent a cost-effective intervention with respect to the standard cost per QALY threshold of £20,000–30,000. However, it is important to consider the uncertainty surrounding this estimate.

The cost-effectiveness plane plots the 10,000 bootstrap replications of incremental cost and QALY estimates to illustrate uncertainty in probabilistic terms (Figure 9). The replications are clustered predominantly in the north-east quadrant, reflecting a positive health gain at an increased cost. Health coaching resulted in an incremental QALY gain in 94% of bootstrap replications and costs were higher than usual care in 69% of replications.

FIGURE 9 Cost-effectiveness plane: full sample with imputed data. Reproduced from Panagioti et al.147 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. (Figure not reproduced here; see the original article.)

The CEAC (Figure 10) demonstrates how the probability that health coaching is cost-effective increases with the decision-maker’s willingness to pay. At the lower bound threshold of £20,000 per QALY, there is a 70.2% probability of health coaching being cost-effective. This rises to 79.0% at the upper bound of £30,000. Compared with usual care, health coaching is likely to be cost-effective in ≥ 50% cases if decision-makers are willing to pay ≥ £8180 for 1 QALY.

FIGURE 10 Cost-effectiveness acceptability curve: full sample with imputed data. Reproduced from Panagioti et al.147 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. (Figure not reproduced here; see the original article.)

Cost-effectiveness analysis: complete-case analysis

Table 45 presents the adjusted estimates of the effect of the offer of health coaching on the incremental costs and QALYs over and above usual care in the complete-case sample.

Health coaching is associated with a mean incremental cost of £497.99 (95% CI –£189.19 to £1185.19). Although higher than the incremental cost associated with health coaching in the full sample with imputation, this difference in cost between arms still does not reach statistical significance. The incremental QALY estimate is also higher, at 0.037 (95% CI 0.0037 to 0.070), indicating that there was a significant QALY gain associated with health coaching among the complete-case sample. This results in an ICER of £13,506.27 per QALY, which is again below the lower bound of recommended cost-effectiveness thresholds.

It is again important to examine the uncertainty around these point estimates in probabilistic terms.

Figure 11 shows that the bootstrapped replications are again clustered predominantly in the north-east quadrant. There is notable shift north above the x-axis and to the right of the y-axis compared with the full sample with imputation, as even more of the replications now fall into this quadrant, representing incremental increases in both costs and QALYs above usual care. Health coaching was more costly in 92.5% of replications and led to an incremental QALY gain in 98.6% of replications. The CEAC illustrates that, at the lower bound (£20,000/QALY), there is a 67.2% probability that health coaching is cost-effective, rising to 82.0% at the upper bound of £30,000 per QALY (Figure 12). Compared with usual care, health coaching is likely to be cost-effective in ≥ 50% cases if decision-makers are willing to pay ≥ £13,570 for 1 QALY.

FIGURE 11 Cost-effectiveness plane: complete-case sample. (Figure not reproduced here; see the original article.)

FIGURE 12 Cost-effectiveness acceptability curve: complete-case sample. (Figure not reproduced here; see the original article.)

Discussion

The economic analysis suggested that health coaching resulted in an incremental increase in both costs and QALYs. When a QALY is valued at £20,000, the primary analysis suggests that there is a 70% probability that health coaching is cost-effective.

We used administrative records to fully capture secondary care utilisation, relying on self-reported data for GP appointments only. This thorough examination of resource utilisation patterns showed that individuals in health coaching had a higher utilisation of planned services and lower use of emergency hospital services than those in usual care.

The cmRCT estimates represent the mean effect of the offer of treatment, rather than the effect of treatment on those receiving treatment. The estimates are diluted further by the use of data collected at fixed time intervals. As the start of treatment varied greatly between individuals, the treatment effect is diluted by the inclusion of this pre-treatment period. The estimates presented therefore represent a very conservative estimate of the impact of health coaching.

A major criticism of conventional trials is that they show the effectiveness of an innovation in a very selected group of patients, but fail to ‘scale’ because of issues such as low rates of acceptability among the wider population and differences between those who take part in trials and those eligible for the intervention.148 Low uptake is not a problem in trials – variable uptake and adherence is inherent in pragmatic trials, by design, on the assumption that utilising interventions in routine settings will be accompanied by these issues, and that the estimates of effect generated are those that are most relevant to NHS decision-makers.149 The PRagmatic Explanatory Continuum Indicator Summary (PRECIS)-2 tool for the assessment of pragmatic trials has a specific rating of the measures in place to ensure adherence, with the highest scores for those with no special measures in place.150

However, conventional pragmatic trials are still selective, as patients are randomised on the basis of their initial willingness to engage with the intervention. In a cmRCT, acceptability in the wider population is built into the design, alongside the usual impact of variable adherence. In such a context, it is very challenging to show benefit, but it was felt that the design was a relevant test of health coaching as applied in an integrated care system as a population health strategy, reaching out to patients assessed as in need, but who may not necessarily be looking for self-management support.