What Is the Potential Value of a Randomized Trial of Different Thresholds to Initiate Invasive Ventilation? A Health Economic Analysis

Abstract

Objectives: To estimate the expected value of undertaking a future randomized controlled trial of thresholds used to initiate invasive ventilation compared with usual care in hypoxemic respiratory failure.

Perspective: Publicly funded healthcare payer.

Setting: Critical care units capable of providing invasive ventilation and unconstrained by resource limitations during usual (nonpandemic) practice.

Methods: We performed a model-based cost-utility estimation with individual-level simulation and value-of-information analysis focused on adults, admitted to critical care, receiving noninvasive oxygen. In the primary scenario, we compared hypothetical threshold A to usual care, where threshold A resulted in increased use of invasive ventilation and improved survival compared with usual care. In the secondary scenario, we compared hypothetical threshold B to usual care, where threshold B resulted in decreased use of invasive ventilation and similar survival compared with usual care. We assumed a willingness-to-pay of 100,000 Canadian dollars (CADs) per quality-adjusted life year.

Results: In the primary scenario, threshold A was cost-effective compared with usual care due to improved hospital survival (78.1% vs. 75.1%), despite more use of invasive ventilation (62% vs. 30%) and higher lifetime costs (86,900 vs. 75,500 CAD). In the secondary scenario, threshold B was cost-effective compared with usual care due to similar survival (74.5% vs. 74.6%) with less use of invasive ventilation (20.2% vs. 27.6%) and lower lifetime costs (71,700 vs. 74,700 CAD). Value-of-information analysis showed that the expected value to Canadian society over 10 years of a 400-person randomized trial comparing a threshold for invasive ventilation to usual care in hypoxemic respiratory failure was 1.35 billion CAD or more in both scenarios.

Conclusions: It would be highly valuable to society to identify thresholds that, in comparison to usual care, either increase survival or reduce invasive ventilation without reducing survival.

Figure 1.

Diagram of health economic model. This diagram shows the model structure, which is the same for both scenarios. An individual begins on the left, at “Person with hypoxemic respiratory failure.” Each individual was simulated through each strategy (usual care or threshold), with the same tree structure for each. The only “decision node” is the first node, where the strategy is chosen. All other nodes are “chance nodes.” The diagram has been simplified using α, β, and γ to denote identical branches of the tree structure. The diagram also shows the two scenarios with their hypothetical thresholds: Primary (early invasive ventilation beneficial, threshold A) and secondary (late invasive ventilation safe, threshold B). IMV = invasive mechanical ventilation.

Figure 2.

What do we mean by the value of information?. This diagram explains value of information and expected value of information comparing two strategies, standard of care and intervention. Positive net benefit favors the intervention, and negative net benefit favors the standard of care. A, We begin with prior knowledge, summarized by a vertical black line showing the expected net benefit. Based on this net benefit, standard of care is preferred. A single study is conducted showing that the intervention is preferred. This study could have had a different result (see right-hand column for distribution of study results). After combining the information from the single study with prior knowledge to make an updated state of knowledge, net benefit is positive so the intervention is preferred. Without the new study, we would have opted for standard of care, and incurred a cost equal to the difference between incremental net benefit in updated knowledge compared with current knowledge (red shaded area). The study has value because it precipitated a change in decision-making. B, We begin with prior knowledge including the distribution of uncertainty around the net benefit. The expected net benefit favors the standard of care. We update the prior knowledge using the information from a new study, but there is a distribution of potential study results (based on prior knowledge). When we combine these distributions, we end up with a new distribution of potential updated knowledge. The opportunity cost of choosing the standard of care (the cost of choosing standard of care when the intervention is actually better) has been reduced, and this is the expected value of information.

Figure 3.

Cost-effectiveness acceptability curve. The proportion of iterations where each strategy has the highest net monetary benefit, vs. willingness to pay, for both scenarios. In the primary scenario, threshold A becomes more cost-effective than usual care at a willingness-to-pay of 75,000 or more per quality-adjusted life year (QALY). In the secondary scenario, threshold B is always slightly more cost-effective. In both scenarios, the difference in cost-effectiveness is small and uncertain (probability of being most cost-effective hovers around 0.5 at willingness-to-pay value of 100,000 Canadian dollars [CADs] per QALY). IMV = invasive mechanical ventilation.

Figure 4.

Population expected value of a randomized trial by sample size. The population expected value of sample information (EVSI, y-axis) for randomized trials of different total sizes (x-axis, logarithmic scale) for both the primary scenario (left) and secondary scenario (right). The EVSI captures the value of the additional information gleaned from a study of a particular size. The value comes from an improved probability of correctly identifying the most cost-effective strategy. The population EVSI is derived by multiplying the per-person EVSI by the annual incidence of eligible people in Canada (5000) and the duration of information relevance (10 yr), discounting future benefits by 1.5% per year. For these values of the EVSI, we used a willingness-to-pay of 100,000 Canadian dollars (CADs). Note that these values only describe value that accrues to the Canadian population.