As patients with CHD have a high risk of a secondary event, there is a need for effective secondary prevention interventions with good uptake. This study examined the cost-effectiveness of a novel telephone-delivered secondary prevention program for MI patients. We found a significant improvement in health status as assessed with the SF-6D in both the HC and UC groups at six and 12 months, although the difference between groups was not significant. The intervention was also associated with higher costs compared to UC. This higher cost was mainly driven by higher non-cardiovascular hospitalisation (e.g. renal dialysis, urinary tract and renal disorders, cancer and gastric disorders) rather than the costs of running the intervention.
The primary outcomes paper reported that the intervention resulted in a significant positive effect on mental HRQoL, as well as Social Functioning and Role Emotional subscales of the SF-36 compared with UC . Patients were also more likely to meet recommended levels of physical activity, BMI, vegetable intake and alcohol consumption. Using the SF-6D summary score, we did not observe an intervention effect on utility at 6 or 12 months. This could be a result of the fact that patients in the HC group had numerically lower scores for physical HRQoL, which may result in no differences on an overall score. It should be noted that UK weights were used for the SF-6D, as no Australian weights are available.
There is very limited data available on the cost-effectiveness of telephone intervention programs as a secondary intervention for patients with following MI. Telephone delivered interventions consist of telephone calls from a health professional who provides health advice (coaching) to encourage and support behavioural changes. In contrast, telemonitoring programs involve patients using monitoring devices to measure certain clinical parameters (e.g. blood glucose levels) of which the results are then transmitted by telephone to the health provider. As such telemonitoring programs are not easily comparable with programs focusing more on behavioural changes (i.e. telephone delivered interventions). A recent Cochrane review included structured telephone interviews or telemonitoring for patients with chronic heart failure, with the latter being less relevant . Nine studies included costing of structured telephone support programs; however, the information available for those studies was limited, with often only information on the cost of the program. As our program was undertaken by a health coach, the costs of each session is relatively low, compared to fully integrated telehealth programs. The cost of the health coaching sessions ($37 per session) accounted for less that 4% of the overall cost for the HC group.
The increased cost for hospitalisation due to causes other than CHD events in the HC group could be due to better education, support and monitoring in this group which may have lead to seeking medical intervention earlier in a disease process. This could potentially result in cost-savings in the longer term. As we have analysed health care utilisation at six months, the longer term health care utilisation and costs is unclear.
Telephonic disease management was not effective and was not cost-effective in a randomised trial of patients with systolic or diastolic heart failure over 18 months . That program improved overall survival; however, the disease management program was costly and did not reduce health service utilisation. Telephone delivered intervention for physical activity and diet in a group of adults with chronic disease was considered cost-effective at an ICER of $12,153/QALY for telephone counselling compared to real-life care ; however that study used a theoretical UC group instead of the group of randomised controls . In another clinical trial comparing telephone counselling with UC reported an ICER of $78,489/QALY . Other telephone-intervention studies have reported cost-savings from reduced readmission rates, reduced length of stay  whereas another study reported an unacceptably high cost per QALY gained of $146,870 . The large variance in cost-effectiveness results has provided impetus, in part, for a large-scale telephone intervention study of health coaching with >45,000 patients .
In the present study there were losses to follow-up and incomplete data; therefore multiple imputation techniques were applied. For hospitalisation rates and costs, data was available for all patients. Therefore, a comparison for the number of hospitalisations between the completed cases and multiple imputation analyses (including all patients) can give an indication of the health of the patients who completed and did not complete the six-month follow-up. Patients who continued follow-up had fewer hospitalisations compared to patients who had missing health services data. No differences in HRQoL were observed between patients with and without missing data, using multiple imputation techniques. While missing data is potentially a limitation of the study, hospitalisation data in public hospitals was available for all patients randomised in the study and the hospitalisation costs accounted for the majority of the difference in cost between the two treatment groups (91% for multiple imputation analysis).
This analysis relied upon self-reported patient data; this leads to some uncertainty around the accuracy of the data. Self-reporting can underestimate the number of GP visits using longer recall periods (one year) compared to one month recall period . We performed an analysis on a subsample of 10% to check patient reported data with claims made by their physician. There was significant underreporting for the six month recall period; however, there was a reasonable correlation between number of reported GP visits and claims made by their GP. As such, linear regression methodology was applied. As we could not verify the self-reported data for visits to specialists, other health professionals, health services and medication use with claims data, the self-reported data has been used without any modification. It is uncertain whether the use of self-reported data would bias the results in favour of health coaching. Hospitalisation data for each patient was available for all public hospital admissions in Queensland; however, some patients reported an admission to hospital in a private hospitals or public hospitals in a different state. Self-reported admission to private hospitals was not different between the two study groups, and therefore no adjustments were performed.