A Markov decision model was constructed to combine the data from published studies, to determine the cost-utility of the interventions. The Markov model cycles were one week in length and a lifetime horizon was adopted, with patients continuing in the model until death or until the model reached a steady state. The authors reported that the analysis was carried out from a societal perspective.
The effectiveness data were identified by a systematic review in the PubMed database. This review was conducted by three authors independently and studies were chosen based on their design and if their results were robust. The main measures of effectiveness were the rate of remnant ablation and thyroid cancer recurrence.
Monetary benefit and utility valuations:
Health-related quality of life data were derived from published literature. The quality of life estimates for pre-ablation, ablation, and post-ablation states were based on published data from the Medical Outcome Study, Short Form (SF) 36 Health Survey. These data were transformed using the SF-6D Health Survey to produce the utility weights.
Measure of benefit:
Quality-adjusted life-years (QALYs) were the measure of benefit and they were discounted at an annual rate of 3%.
The direct costs included physician office visits, tests, and surgery, and they were from the Medicare reimbursement schedule. The productivity costs, arising from the loss of working days, were from the US Bureau of Labor Statistics. The price year was 2009. The costs were in US dollars ($) and they were discounted at an annual rate of 3%.
Analysis of uncertainty:
A one-way sensitivity analysis was performed on all the variables that differed between the two interventions. All the inputs of the model were varied by reasonable ranges from the literature. If no range was found in the literature review, the inputs were varied by ±5 to 10% of their base-case estimate. A 95% confidence interval for the inputs was calculated using Monte Carlo simulations. The results were presented in a tornado diagram and in line graphs.