Background: The cost-effectiveness of a heart failure management intervention can be further informed by incorporating the expected benefits and costs of future survival.

Methods: This study compared the long-term costs per quality-adjusted life year (QALY) gained from home-based (HBI) vs specialist clinic-based intervention (CBI) among elderly patients (mean age = 71 years) with heart failure discharged home (mean intervention duration = 12 months). Cost-utility analysis was conducted from a government-funded health system perspective. A Markov cohort model was used to simulate disease progression over 15 years based on initial data from a randomized clinical trial (the WHICH? study). Time-dependent hazard functions were modeled using the Weibull function, and this was compared against an alternative model where the hazard was assumed to be constant over time. Deterministic and probabilistic sensitivity analyses were conducted to identify the key drivers of cost-effectiveness and quantify uncertainty in the results.

Results: During the trial, mortality was the highest within 30 days of discharge and decreased thereafter in both groups, although the declining rate of mortality was slower in CBI than HBI. At 15 years (extrapolated), HBI was associated with slightly better health outcomes (mean of 0.59 QALYs gained) and mean additional costs of AU$13,876 per patient. The incremental cost-utility ratio and the incremental net monetary benefit (vs CBI) were AU$23,352 per QALY gained and AU$15,835, respectively. The uncertainty was driven by variability in the costs and probabilities of readmissions. Probabilistic sensitivity analysis showed HBI had a 68% probability of being cost-effective at a willingness-to-pay threshold of AU$50,000 per QALY.

Conclusion: Compared with CBI (outpatient specialized HF clinic-based intervention), HBI (home-based predominantly, but not exclusively) could potentially be cost-effective over the long-term in elderly patients with heart failure at a willingness-to-pay threshold of AU$50,000/QALY, albeit with large uncertainty.