Contrast Echocardiography Cost Effectiveness for Left Ventricular Thrombus Assessment

Decision analysis techniques will be utilized to construct and estimate a model for the diagnosis, treatment, outcomes and costs among patients with dilated or ischemic cardiomyopathy referred for clinically indicated rest transthoracic echocardiography. We will first perform a systematic review of the literature regarding the use of contrast echocardiography for detection of LV thrombus, in order to ensure that model design accurately reflects clinical practice and current outcomes. Patients will begin the model in a state of dilated or ischemic cardiomyopathy with possible LV thrombus, and will progress through the model's various health states based on event probabilities that will be influenced by whether the patient received contrast during their initial echocardiogram. The probability of events occurring (e.g., stroke, bleeding complication, death) will be estimated using data from the literature and the DEDUCE database. The cost of inpatient, outpatient, and extended care for health states will be derived from available sources, including Medical Expenditure Panel Survey data, National Nursing Home Survey data, and published literature. Appropriate statistical methods, including inverse probability weighted partitioned estimators, will be used to correctly impute long-term resource utilization, given that the data include patients who are censored due to loss to follow-up. Quality of life adjustment factors for health states will be drawn from the literature. For temporary event states such as bleeding, a utility decrement will be subtracted from the baseline utility value.

The model will have a long-term horizon, and model outcomes will include long-term costs as well as QALYs. There will be a number of scenario analyses to evaluate alternative modeling assumptions. For example, we will simulate the selective use of contrast in only those patients with a technically inadequate study versus the routine use of contrast in the primary analysis. We will also evaluate cost-effectiveness from alternative perspectives (U.S. health care system; provider; payer), under various reimbursement schemes of interest (e.g., fee-for-service; accountable care organization setting) and assuming short term time horizons (e.g., index episode of care through 90 days after testing; three year horizon). We will also test the model's robustness through a series of sensitivity analyses. These will include univariate (one-way) sensitivity analyses in which we will assess the model's robustness to alternative values of each input parameter individually. Examples of parameters for which sensitivity analyses will be conducted include costs, thrombus incidence, and event rates. In addition to univariate sensitivity analyses, we will also perform probabilistic (many-way) sensitivity analyses in which we use Monte Carlo methods to assess the model's global stability. Cost-effectiveness acceptability frontiers will be generated to illustrate model sensitivity. All modeling will be performed in a manner consistent with current International Society for Pharmacoeconomics and Outcomes Research Good Research Practices guidance. In order to incorporate variability in willingness-to-pay thresholds in United States payers, we will report our results across a range of cost-effectiveness thresholds from $50,000 - $200,000 per QALY, with $100,000 per QALY pre-specified as the primary threshold.