Cost-Effectiveness Analysis in Cardiac Surgery: a Review of Its Concepts and Methodologies

Abstract

Cost-effectiveness analysis (CEA) in cardiac surgery continues to grow in relevance with increasing health care expenditures, a greater emphasis on value-based care, the continuing development of costly surgical and noninvasive technologies, advances in cardiac devices, and changes in eligibility criteria over the past two decades. Although the rapidly evolving surgical technologies pose challenges to CEA, improvements in gathering and leveraging long-term economic and clinical data alongside trials and in cardiac surgery registries represent future opportunities for the field. As such, it is important for cardiac surgeons to understand CEA with respect to existing and future surgical therapies. Herein, we review the fundamental principles of cost-effectiveness analysis theory and discuss recent cost-effectiveness studies on cardiac surgery.

INTRODUCTION

Over 80 million adults in the United States (US) suffer from some form of cardiovascular disease, accounting for close to one in three US deaths annually and over $300 billion in direct and indirect costs.¹ Coronary heart disease has been estimated to affect over 6% of the US adult population. Moderate to severe aortic stenosis (AS) and mitral regurgitation (MR) have been estimated to affect close to 3% and 9% of US adults ages 75 and older, respectively.^{1, 2} Atrial fibrillation (AF) and heart failure (HF) each affect up to six million Americans.¹ The development of new and improved technologies, including minimally invasive and hybrid revascularization procedures, transcatheter aortic valve replacement (TAVR), MitraClip, continuous-flow left ventricular assist devices (LVADs), and ablation devices for AF, has greatly changed our approach to these conditions and expanded indications for treatment.³

With increasing health care expenditures,⁴ and a health policy environment promoting greater efficiency and value-based care,⁵ the relevance of evaluating cost-effectiveness in cardiac surgery has become more critical. The growing focus on cost-effectiveness research in cardiac surgery can be shown by an increasing number of publications in the field (Figure 1). While a portion of this trend may be the result of the aforementioned health care system factors, the continuously changing surgical landscape with approval of new devices has also been an enabler of CEA in cardiac surgery. For example, there has been a steady increase in the number of CEA publications that have focused on the treatment of AS and MR since 2011, and almost all have evaluated new procedures such as TAVR and MitraClip (Figure 1). Though other countries have adopted CEAs into their budgetary considerations, U.S. federal payers have not explicitly used CEA to establish guidelines and costs have only been considered implicitly.^{4, 6-8} However, more recently, the ACC/AHA has recommended the inclusion of CEA in their clinical guidelines, while other public and private sector organizations have also incorporated value-based measures in their analyses.^{4, 7, 9, 10}

FIGURE 1. Cost-effectiveness analyses in cardiac surgery published since January 2000.

LVAD, left ventricular assist device; DT, destination therapy; DES, drug-eluting stent; TAVR, transcatheter aortic valve replacement; CF, continuous-flow; ACA, Affordable Care Act; BTT, bridge to transplant; AS, aortic stenosis; MR, mitral regurgitation.

Clinicians, as well as payers, are critical for effectively and efficiently allocating society’s health care resources and maximizing value through evidence-based decisions. Although health care economics education has been increasingly incorporated into the standard medical school curriculum for physicians in training, it may not be sufficient.^{4, 11, 12} As such, this paper was written as a primer on the theory and application of cost-effectiveness analysis for cardiac surgeons. We additionally summarized the findings from recent CEAs on five cardiac conditions: CAD, AS, MR, AF, and end-stage HF, with a focus on the latter to illustrate the use of CEA for guiding surgical decision-making.

METHODS

We developed a PubMed search for CEAs published since January 2000 and in English language, evaluating cardiac surgical interventions for management of these five cardiac conditions. Search terms included combinations of MeSH terms and keyword variations for CABG, aortic valve replacement, mitral valve surgery, surgical ablation, MAZE, LVAD, and the applicable cardiac conditions. To capture CEAs, MeSH terms and variations of “cost-effectiveness analysis” and “quality-adjusted life years” were combined with the aforementioned search terms.

Articles were selected based on a review of titles and abstracts followed by a text review. We only included analyses with both cost and effectiveness components. The effectiveness component was limited to quality-adjusted life years (QALY) or life-years (LY). We selected 65 articles in which the analysis included at least one cardiac surgical intervention (Figure E1).

We extracted all of the relevant information from the CEAs and developed matrices, grouped by the five conditions. For each matrix, we delineated the target population, setting and location, comparisons made, time horizon, and base case measures of cost-effectiveness. Table 1 depicts the matrix for end-stage HF and Tables E1-4 depict the matrices for the four other conditions.

TABLE 1.

Summary of cost-effectiveness analysis findings for end-stage heart failure

Author	Year	Comparison	Country	Horizon	Cost year	ΔCosts	ΔEffectiveness*	ICER*
LVAD Destination Therapy
Samson44	2004	Pulsatile LVAD vs. MM	USA	Lifetime	2002	$338,882	0.42	$802,700
Clegg45	2007	Pulsatile LVAD vs. MM	UK	5 years	2003	£101,998	0.59	£170,616
Rogers27	2012	CF LVAD vs. MM	USA	5 years	2009	$297,551	1.5	$198,184
Neyt32	2013	CF LVAD vs. MM	Dutch	Lifetime	2010	€299,100	2.83	€107,600
Long46	2014	CF LVAD vs. MM	USA	Lifetime	2012	$480,400	2.38	$201,600
Baras Shreibati47	2016	LVAD vs. MM	USA	Lifetime	2016	$364,400	1.74	$209,400
LVAD Bridge-to-Transplant
Clegg48	2006	Pulsatile LVAD vs. MM	UK	5 years	2003	£99,475	1.53	£65,242
Sharples49	2006	CF/Pulsatile LVAD vs. MM	UK	Lifetime	2004/05	£42,936	−1.72	LVAD dominated by MM
Moreno50	2012	CF LVAD vs. MM	UK	Lifetime	2011	£142,495	0.55	£258,922
Alba51	2013	CF LVAD vs. MM**	Canada	20 years	2011	$100,841	1.19 (LY)	$84,964 (/LY)
		CF LVAD vs. MM**	Canada	20 years	2011	$112,779	1.14 (LY)	$99,039 (/LY)
		CF LVAD vs. MM**	Canada	20 years	2011	$144,334	1.21 (LY)	$119,574 (/LY)
Sutcliffe52	2013	CF LVAD vs. MM	UK	Lifetime	2010	£135,726	2.46	£55,173
		CF LVAD ATT vs. CF LVAD	UK	Lifetime	2010	−£32,813	−1.59	£20,637
Clarke43	2014	CF LVAD vs. MM	UK	Lifetime	2011	£127,391	2.38	£53,527
Pulikottil-Jacob53	2014	HeartWare CF LVAD vs. HeartMate II CF LVAD	UK	Lifetime	2011	£27,042	1.14	£23,530
Long46	2014	CF LVAD vs. MM vs. no transplant	USA	Lifetime	2012	CF LVAD vs. MM: $482,900; MM vs. no transplant: $398,700	CF LVAD vs. MM: 2.13; MM vs. no transplant: 4.12	CF LVAD vs. MM: $226,300; MM vs. no transplant: $96,900

Abbreviations: CF, continuous-flow; LVAD, left-ventricular assist device; LY, life-years; MM, medical management.

^*ΔEffectiveness and ICERs were calculated using QALYs unless specified to be life-years.

^**High, medium and low risk from top to bottom.

COST-EFFECTIVENESS ANALYSIS

Within a formal CEA, the average costs, in currency units, and health outcomes of the relevant competing medical options can be compared for a particular patient, e.g. the “average” or typical patient, or a heterogeneous population. Health outcome (the measure of effectiveness) is preferably expressed as life expectancy adjusted for time spent at less than full quality, i.e. “quality-adjusted life expectancy”, typically measured in quality-adjusted life years (QALYs) (Figure 2). Generally, CEAs are pragmatic in that they evaluate and compare the effects of medical options on costs and health outcomes in the setting of usual clinical practice. Although many of the CEAs we identified compared just two treatment options, in instances where there are greater than two relevant treatment options, all should be considered in the analysis.

FIGURE 2: Hypothetical individual patient’s follow-up duration adjusted for quality-of-life.

The patient’s health state is longitudinally measured using a health state classification instrument at preoperative and several postoperative time points. The health states are then converted into utilities using HRQoL weights based on societal preferences. Quality-adjusted life years (QALYs) are represented by the area under the curve, i.e. the sum of each period multiplied by the HRQoL/utility during that period. Zero indicates death while 1 indicates perfect health. HRQoL, health-related quality-of-life; QALY, quality-adjusted life year.

Once the average cost and effectiveness of all of the relevant alternative options are measured, one can then order them by cost, from lowest to highest. Any option that costs the same or more than a competing option but is less effective is clearly less desirable and should be rejected from further consideration. Such options are said to be dominated. The options that remain, i.e., those that are not eliminated due to dominance, are now in order of both increasing costs and increasing effectiveness, and can be compared two at a time to determine whether the added cost of the more expensive and more effective option in the pair meets our expectation of good value. The metric used for estimating value is the incremental cost-effectiveness ratio (ICER), which is measured in costs per additional unit of health gained, and is calculated as the difference in average costs of the two options under consideration divided by the difference in their average effectiveness, i.e. $\frac{{\overline{\text{Costs}}}_1-\overline{{\overline{\text{Costs}}}_2}}{{\overline{\text{QALYs}}}_1-{\overline{\text{QALYs}}}_2}$. Cost-effectiveness is then assessed in pairs for remaining options by comparing the ICER for each pair with a cost-effectiveness threshold value, i.e. the presumed maximum dollar amount that society would be willing to pay for a gain in a unit of health. For the US, there is currently no single agreed upon cost-effectiveness threshold but measures in the range of $50,000 to $200,000 per QALY have been used and recommended.⁴ Conclusions from a CEA about implementing interventions are based on the mean cost and QALY estimates, irrespective of their uncertainty. Uncertainty around the estimates is more relevant to deciding whether also further research is required. Table 2 illustrates these principles comparing three treatment options for patients with end-stage HF (medical management, axial-flow LVAD, and centrifugal-flow LVAD).

TABLE 2.

Hypothetical examples of CEAs comparing centrifugal continuous-flow LVAD vs. axial continuous-flow LVAD vs. medical management

Scenarios*	Costs ($)	QALYs	ICER ($/QALY)	Comparison
Scenario 1 – Dominance
Medical management	53,000	0.41	N/A	N/A
Centrifugal LVAD	390,000	1.92	223,179	Centrifugal LVAD vs. Medical management
Axial LVAD	416,000	1.88	Dominated by centrifugal LVAD	Axial LVAD vs.centrifugal LVAD
Scenario 2 – Extended dominance
Medical management	53,000	0.41	N/A	N/A
Axial LVAD	392,000	1.91	(226,000) Extended dominance	Axial LVAD vs. Medical management
Centrifugal LVAD	406,000	1.99	(175,000)	Centrifugal LVAD vs. axial LVAD
			223,418	Centrifugal LVAD vs. Medical management
Scenario 3 – One ICER found to be below cost-effectiveness threshold of $200,000/QALY
Medical management	53,000	0.41	N/A	N/A
Centrifugal LVAD	252,000	2.01	124,375	Centrifugal LVAD vs. Medical management
Axial LVAD	392,000	1.91	Dominated by centrifugal LVAD	Axial LVAD vs. centrifugal LVAD
Scenario 4 – Two ICERs found to be below cost-effectiveness threshold of $200,000/QALY
Medical management	53,000	0.33	N/A	N/A
Axial LVAD	301,000	2.33	124,000	Axial LVAD vs. Medical management
Centrifugal LVAD	352,100	2.7	137,838	Centrifugal LVAD vs. axial LVAD

Abbreviations: CEA, cost-effectiveness analysis; ICER, incremental cost-effectiveness ratio; LVAD, left ventricular assist device; QALY, quality-adjusted life year.

In scenario 1, axial LVAD is dominated by centrifugal LVAD because, on average, it both costs more and is the least effective of the two. Therefore, it should be eliminated from further consideration. The ICER that compares the two remaining treatment options (centrifugal LVAD and medical management) is $223,379, which is above the proposed cost-effectiveness threshold of $200,000/QALY. Consequently, in this scenario, medical management is the most cost-effective option for treating advanced heart failure.

Scenario 2 illustrates the concept of extended dominance. Centrifugal LVAD both costs more and is more effective than the axial flow LVAD option. However, what we observe here is that the cost per each additional unit of health gained by centrifugal flow LVAD therapy over axial flow LVAD therapy ($175,000) is less than the cost for each additional unit of health gained with axial flow LVAD over medical management ($226,000). It follows that centrifugal flow LVAD will generate health at a rate cheaper than axial flow therapy ($223,418) and, therefore should be the preferred option. So, by “extended dominance”, axial flow therapy is eliminated. However, because the ICER for centrifugal therapy compared to medical therapy is over the $200,000/QALY threshold, medical management is likely the most cost-effective option for treating advanced heart failure in this scenario.

In scenario 3, axial LVAD is dominated by centrifugal LVAD and the ICER for centrifugal LVAD vs. medical management is below the cost-effectiveness threshold, indicating centrifugal LVAD is the best option.

In scenario 4, although both ICERs lie below the cost-effectiveness threshold, centrifugal LVAD offers the greatest overall health benefit and is considered the most attractive option here.

^*The values provided have been loosely adapted from actual studies and serve as illustrations.

ESTIMATION OF COSTS

A key component of CEA is calculating the average cost of each alternative intervention. Typically, CEAs are conducted from a health care or societal perspective. CEAs from a health care perspective should capture current and future formal health care costs including those incurred by third party payers and patients’ out of pocket expenses.⁸ Formal health care costs include those directly and indirectly related to the disease or its management.^{8, 13, 14} Typically, for surgical interventions, it includes costs related to the index procedure, additional hospitalizations, physician fees, and other costs, such as rehabilitation facilities, nursing homes and outpatient care.

When a societal perspective is chosen, informal health care costs, including those associated with patients’ time, care from family members or others that were not reimbursed, and transportation should be incorporated, in addition to formal health care costs. Non-health care costs, such as those associated with lost productivity, non-medical consumption and other impacted items may also be included.^{8, 13, 15}

Medical resource use and its associated costs can be gathered prospectively in the setting of randomized controlled trials (RCT), observational studies or taken from multiple secondary sources. Total costs are often not captured directly, but rather approximated by multiplying resource use (e.g., medical personnel hours) by unit costs or by applying CMS established relative price weights, such as those derived from diagnostic related groups (DRGs) and fee schedules.^{14, 16-18} Because such prospective payments are based on average resource use this leads to loss of cost variability and precision across patients.^17-20 In some instances, patient level claims data are used, which capture both resource use and associated charges,^{13-16, 21} however, such charges need to be converted to costs (i.e., the actual value of the resources consumed), which can be accomplished using institution specific cost-to-charge ratios (obtainable from CMS hospital cost reports).^{13-16, 21-23} When charges are available at a cost center or department level (e.g. from uniform billing forms), departmental level cost-to-charge ratios can be utilized.^{17, 18, 24-26} However, when only total hospitalization charges are available, hospital-level aggregated cost-to-charge ratios may be used. There are times when resource use and costs are not gathered at all and expected costs derived from similar studies may be used as a proxy, conditional on the occurrence of clinical outcomes.^{8, 13, 15, 16} An example of some of these methods can be found in a CEA comparing long-term continuous-flow LVAD therapy to medical management, which utilized hospital claims for LVAD patients enrolled in a RCT to estimate procedure costs, Medicare claims data to estimate physician fees, DRG-based Medicare reimbursement rates to estimate re-hospitalization costs, and a study of bridge-to-transplant (BTT) LVAD patients as the source for estimates of outpatient costs.²⁷

ESTIMATION OF EFFECTIVENESS

Formal CEAs integrate a valuation of health with survival to generate a composite measure of effectiveness, quality-adjusted life expectancy, for each intervention being compared. Analyzing unadjusted life expectancy as well helps to demonstrate the extent to which analytical results are influenced by non-fatal vs. fatal events.^27-33 Health states can be measured longitudinally by periodically administering health status instruments to patients, e.g. before and after an intervention. Commonly used generic, health-related quality-of-life (HRQoL) indexes, which are also suitable for use in cardiac surgery patients, include the EuroQoL, Health Utilities Index, SF-36, and SF-12.^{15, 16, 34, 35}

For CEAs conducted from a societal or health care perspective, patient responses to these generic HRQoL indexes can then be valued according to community preferences determined by a sample of the general population.^{8, 15, 16, 36} Applying such community preference weights to health states transforms them into a utility score, with higher values indicating greater well-being.^{15, 16, 36} On a utility scale, the best possible health state, i.e. full health, is assigned a value of one, while death is assigned a value of zero. Some instruments, however, depict health states perceived as being worse than death by utility scores below zero.

For a given individual, quality-adjusted life expectancy is calculated by multiplying the duration of each time period with a consistent quality-of-life by the associated quality-of-life preference measure (i.e., utility score) and then summing all utility-weighted periods over the entire time horizon (Figure 2).^{4, 15, 16} For example, a life expectancy of 10 years at a utility preference weight of 0.5 is equivalent to five years lived at a full health (utility preference weight of one). Typically, quality-adjusted life expectancy is expressed in quality-adjusted life years (QALY). When comparing average quality-adjusted life expectancy for alternative interventions, one calculates the difference in the “shaded” area of Figure 2 across all patients per intervention.

Sometimes HRQoL metrics may not have been collected in the course of a study. In that case, one would need to use proxy values derived from the literature for the utility weighting of health states. For example, the previously mentioned CEA on continuous-flow LVADs derived the needed HRQoL adjustments from published reports, dependent on NYHA class, rather than measuring them directly in the patients who were under study.²⁷

STUDY DESIGN AND CONSIDERATIONS

Understanding the findings of a CEA requires knowledge of the quality of the data sources, of its perspective (e.g., societal or payer), of its time horizon, and whether outcomes were modeled or measured. For model-based CEAs, model type, parameter assumptions, and model validity are important considerations.

CEAs may just utilize data from an RCT, referred to as within-trial CEAs, but will often develop a decision model that integrates various data types (survival, morbidity and quality-of-life). Different decision modeling methods exist, but the most frequently used technique is state-transition modeling (e.g. Markov) based on pre-defined health states, rates of mortality and other events.^{15, 37, 38} For example, in a CEA using a Markov model that compared PCI stenting with minimally invasive CABG for LAD disease, probabilities for repeat revascularizations and adverse events were derived from a meta-analysis and other literature sources.³⁹

CEAs may derive results based solely on the observed period for which patient data was collected or, may report conclusions based on extrapolated future outcomes. While projections beyond the observed data require assumptions, strict within-trial data may be too short-term to provide a meaningful estimate of cost-effectiveness. For example, a CEA of ablation surgery for AF that used one-year within-trial data only, found that concomitant ablation was too costly due to longer operation time and catheter ablations costs, i.e., the full benefits of performing this procedure were not adequately captured during the relatively short study follow-up period.⁴⁰ However, another analysis, which utilized a decision model that projected 15 year outcomes by assigning probabilities to adverse events and survival beyond the period of time for which there were empirical data, was able to demonstrate ablation’s cost-effectiveness.⁴¹ When CEAs evaluate outcomes over an extended time horizon covering multiple years, they should also adjust for time preference, i.e., the fact that individuals typically value years of life experienced in the nearer future more dearly than those experienced later in life. This is analogous to how we value dollars we can spend now more than an equivalent amount we could spend in the future. Typically, both future costs and health outcomes are adjusted downward by using a discount rate (recommended at 3% for US).^{8, 15, 16}

To increase credibility of the model’s predictions, model performance should be evaluated. With external validation, model predictions, such as survival, are compared with independently measured data from another trial or observational studies.

UNCERTAINTY AND SENSITIVITY ANALYSIS

Uncertainty analysis is conducted to quantify the impact that a range of plausible cost and effectiveness input values could have on the model’s outcomes and related recommendations. When CEA results are uncertain, one may want to recommend further research to obtain more information. A relevant source of uncertainty in cost-effectiveness analysis are the parameter estimates, i.e., uncertainty related to clinical event rates, utility scores, costs and other model inputs. Such uncertainty arises from the size and variability of the study from which the data was derived and the validity and generalizability of that study. Therefore, rather than only using input values based on averages, CEAs ideally utilize a range of plausible input values that give rise to different cost-effectiveness outcomes.^{15, 42}

Calculating a range of outcomes due to parameter uncertainty is performed through “probabilistic” or “deterministic sensitivity analysis”. In model-based CEAs with probabilistic sensitivity analysis (PSA), different combinations of all input values are randomly selected from a priori defined parameter distributions and for each set of parameter values the model is run to produce a distribution of cost-effectiveness outcomes. When using patient-level data (e.g. in within-trial CEAs), bootstrapping can be performed without pre-specifying input distributions. Results from PSA can be summarized as 95% confidence intervals (95% CIs) around cost and effectiveness outcomes or as the percent of bootstrap iterations in which a particular intervention is the most cost-effective option given a chosen cost-effectiveness threshold (Figures 3,4).^{15, 42} One CEA showed that higher cost-effectiveness thresholds increased the probability of BTT-LVAD’s cost-effectiveness, e.g. given a cost-effectiveness threshold of £50,000/QALY, BTT-LVAD would be economically attractive in 41% of the PSA iterations.⁴³ In deterministic sensitivity analysis, individual parameter values are varied by the researcher within a realistic range to test how they impact outcomes (Figure 5).^{15, 42}

FIGURE 3: Hypothetical probabilistic sensitivity analysis (PSA) plotted on a cost-effectiveness plane.

When comparing two competing surgeries, A versus B, a scatterplot of the difference in average costs and QALYs per PSA iteration can be created with a diagonal representing the cost-effectiveness threshold. The percentage of points lying to the right of a given threshold line indicates the probability that the intervention is cost-effective relative to the competing intervention. Multiple cost-effectiveness thresholds can be plotted to determine the impact on the probability of cost-effectiveness. The lower right quadrant represents iterations where the intervention A is “dominant” due to having lower incremental costs and higher incremental QALYs than B. The upper left quadrant represents iterations where A is “dominated” due to higher incremental costs and lower incremental QALYs. The upper right and lower left quadrants represent tradeoffs between higher and lower incremental costs and QALYs, respectively.

QALY, quality-adjusted life year.

FIGURE 4: Cost-effectiveness acceptability curves.

This graph shows the probability of each intervention being cost-effective given a range for society’s willingness to pay to gain one QALY. As the cost-effectiveness threshold increases, the probability that surgery A is cost-effective increases while that of B decreases (equal to 100% - probability A is cost-effective). The vertical lines represent just two of the cost-effectiveness thresholds and correspond directly to the diagonals on the cost-effectiveness plane.

QALY, quality-adjusted life year.

FIGURE 5: One-way deterministic sensitivity analyses across several model inputs.

The base case scenario represents the incremental cost-effectiveness ratio (ICER) point estimate when comparing two surgeries. In deterministic sensitivity analysis, a given input, e.g. the HRQoL weight, is varied in the model to determine how upper and lower bound assumptions impact outcomes. For example, when comparing surgery A versus B, assuming a higher HRQoL following A lowers the ICER, as incremental QALYs increase.

QALY, quality-adjusted life year; HRQoL, health-related quality-of-life.

COST-EFFECTIVENESS OF LVAD THERAPY FOR END-STAGE HEART FAILURE

To illustrate the value of CEA, we reviewed the CEA literature on LVAD therapy for end-stage HF. Thirteen CEAs were identified by our literature search, of which six studies compared LVAD as destination therapy (DT) with medical management in heart transplant ineligible patients,^{27, 32, 44, 45, 46, 47} seven studies compared BTT-LVAD with medical management in transplant eligible patients,^{43, 46, 48-52} and one study compared second to third generation BTT-LVADs (Table 1).⁵³

For DT, all generations of LVADs were consistently more costly than medical management, mainly due to the high upfront implantation costs and costs associated with re-hospitalizations. However, all studies found that LVADs improved survival and quality-of-life (QoL). CEAs conducted for the US health care system demonstrated a substantial improvement in the ICER over time, which can be mainly attributed to improved survival, reduced implantation costs, improved patient selection, and reduced device complications observed with newer generation LVADs.^{27, 44, 46, 47} The effectiveness of LVAD increased by 0.42-0.59 QALY with pulsatile and 1.5-2.83 QALYs with continuous-flow device technology. A direct comparison of cost estimates among CEAs on DT-LVAD vs medical management remains difficult however, as the methods and sources for costs differed across studies: for estimation of inpatient costs charges were converted into costs^{27, 44, 46} and/or payment data from fee-for-service Medicare beneficiaries were used.^{27, 47}

Studies on BTT-LVADs showed a wide variation in improvement of QALYs when medical management was the comparator, sometimes resulting in very different conclusions on cost-effectiveness. For example, one study showed that LVAD therapy was both more costly and less effective than medical management (i.e., dominated).⁴⁹ However, this particular study assumed that patients with LVADs would receive transplants much later than those on medication and that mortality rates became equal across treatment groups during the “bridged period”, both assumptions that are unjustified. All other studies assumed similar transplant rates across treatment groups and lower mortality with BTT-LVAD.^{43, 46, 48, 50-52} The study with low incremental effectiveness (0.55 QALY),⁵⁰ used relatively short time to transplant estimates: an average waiting time of 6 months. In this study, BTT-LVADs were found to become much more cost-effective when assuming a longer time to transplant (18 months): the ICER dropped from £258,922 to £133,860. Time to transplant in other analyses was assumed to be much longer though (median time ~45 months), potentially explaining the apparently larger gain of ~2.4 QALYs and lower ICERs around £55K.^{43, 52}

The variation in CEA findings for BTT-LVAD when medical management was the reference might be further explained by the lack of randomized trial data in the BTT realm. Therefore, survival rates during the “bridged” period had to be based on retrospective cohort series^{48, 49, 51} or on separate analyses of treatment arms in organ transplant and VAD registries, impeding appropriate confounder adjustment.^{43, 46, 50, 52}

For DT-LVAD, trial data were used to model survival,^{44, 45} although CEAs comparing continuous-flow DT-LVAD with medical management were based on an indirect comparison from RCTs on pulsatile LVAD vs medical management and continuous-flow vs pulsatile LVAD.^{27, 32} Other CEAs used data from VAD registries for the latter comparison.^{46, 47}

The question arises whether the findings from CEAs of LVAD vs medical management are in agreement with clinical practice guidelines endorsing the use of both DT- and BTT-LVADs in end-stage HF (class IIa recommendation).^54-57 CEAs in recent years show ICERs that can be considered borderline acceptable, especially at a generally higher societal willingness-to-pay in the context of end-of-life care.^{58, 59} Because it is difficult to use economic arguments for contesting the current practice of LVAD, a procedure that has been shown to save lives and improve quality-of-life, it may become more relevant to evaluate the use of different next generation LVADs, such as HeartWare and HM III within CEA.⁵³ Recently two RCTs were published comparing these newer centrifugal-flow devices with the existing axial-flow LVADs, as DT⁶⁰ and DT/BTT.⁶¹ Both trials showed that centrifugal-flow LVADs have similar survival and are associated with lower device failure rates, although in transplant ineligible patients they may result in higher stroke rates.⁶⁰

Beginning with the seminal REMATCH trial,⁶² LVADs have served as one of the most highly studied modern cardiac surgical devices, transforming the contemporary management of end-stage HF. LVADs represent an effective yet costly therapy with significant variability in cost-effectiveness outcomes across studies. Device experience has driven device innovation and improved patient selection which, in turn, has resulted in improved clinical outcomes and ICERs. While CEA findings do not currently enter into formal consensus guideline recommendations in the US, they are nevertheless critical for understanding the balance between the clinical need of advanced HF therapy with availability of resources,⁵⁹ and serving as benchmarks for next generation devices.

DISCUSSION

The importance of integrating CEA into decision-making in cardiac surgery continues to grow with the greater emphasis on value-based care, and the development of novel devices and procedures.^63-68 With constrained resources, payers will increasingly take into account value (i.e. cost, per unit outcome) in making coverage and reimbursement decisions regarding new surgical interventions.^{11, 12} Moreover, with the movement towards population health management --where health care systems are responsible for the long-term health outcomes and costs of the populations they serve-- economic analyses have become increasingly relevant for clinical decision making.

At the same time, analyzing the cost-effectiveness of cardiac surgery poses some methodological challenges. Often cardiac surgery interventions have high upfront costs and risks that may be off-set by long-term gains in survival, quality-of-life, and reductions in morbidity and health care resource utilization. As such, the selection of a study’s time horizon can substantially impact the results.^{40, 41, 69}

Another challenge is related to the innovative nature of cardiac surgery, which may include the development or incremental modification of devices or ongoing changes in surgical technique, patient selection, and peri-operative management of patients.⁷⁰ As such, CEAs need to address the “moving target phenomenon,” either by incorporating potential changes into sensitivity analyses or by planned reassessments. For example, the improvements of axial and continuous-flow LVADs over pulsatile LVADs have necessitated updated CEAs, with substantial improvements in its ICER relative to medical management.⁵⁹ Now that DT and BTT-LVADs have been widely supported by clinical guideline societies and with the approval of centrifugal-flow device technology, comparative CEA of currently available devices becomes more relevant.

The usefulness of CEA depends heavily on the quality of the underlying data and assumptions for synthesis and extrapolation of the evidence selected. Care delivered and the patients who participate in research studies may not be representative for the current practice, limiting generalizability of findings. In the absence of robust long-term follow-up data on both clinical and economic outcomes and good cost estimation methodology, CEAs on the same topic may vary, even when the perspective and setting is similar, as shown for the CEAs concerning LVADs. Criteria for a useful CEA are summarized in Table E5.

Fortunately, the ability to generate data to conduct CEAs is improving. One important development has been the investment made by hospitals and large health systems in electronic health records and cost data, which are increasingly accessible through data warehouses.⁷¹ Moreover, economic data is increasingly recognized by research funding agencies as an important component of research (trials and other prospective studies) to evaluate new interventions.¹³ Linkages between registries, such as the STS National Database,⁷² and national databases, such as the Medicare database, represent opportunities to track longer term outcomes and health resource use. In addition to its other databases, STS has established the STS/ACC TVT Registry to track 30-day and one-year outcomes for institutions conducting transcatheter aortic and mitral-valve repair or replacement operations.^{68, 73} Estimation of costs could however be improved by drawing on more detailed and accurate internal cost-measuring systems adopted by many US hospitals instead of relying on indirect estimation using charges.⁷⁴

Traditionally, most CEAs have been designed to give answers for the “average” patient or patient population as a whole, receiving “average” care. However, when treatment effects are heterogeneous, approaching decision problems from such a “one-size fits all” perspective will lead to suboptimal outcomes in patient subgroups. Yet, in an era of increasing individualization of care, there is a higher demand for CEAs that also provide results applicable to the individual patient. For example, when older age and higher comorbidity are associated with higher immediate surgical risks and costs, less invasive treatments may become more attractive, especially when the downstream benefits with surgery are foreseen to get minimized by the patient’s limited life expectancy. Because CEAs aim to integrate all potential harms and benefits within the analysis, individualized CEAs are uniquely positioned to improve patient selection and guide personalized medical decision-making, further optimizing value of care.⁷⁵

Understanding the methods underlying cost-effectiveness analysis is critical in this environment of constrained resources and ongoing policy changes that affect financial incentives in order to ensure clinical participation in further shaping the health care system.⁷⁶

Extended Data

TABLE E1.

Summary of cost-effectiveness analysis findings for aortic stenosis

Author	Year	Comparison	Country	Horizon	Cost year	ΔCosts	ΔEffectiveness*	ICER*
Inoperable
Gada1	2012	TA TAVR vs. MM	USA	Lifetime	2012	NR	NR	$44,384
Gada2	2012	TF TAVR vs. MM	USA	Lifetime	2011	NR	NR	$39,964
Neyt3	2012	TF TAVR vs. MM	Belgium	Lifetime	NR	€33,200	0.74	€44,900
Reynolds4	2012	TF TAVR vs. MM	USA	Lifetime	2010	$79,837	1.29	$61,889
Watt5	2012	TF TAVR vs. MM	UK	10 years	2010	£25,200	1.56	£16,200
Doble6	2013	TF TAVR vs. MM	Canada	20 years	2010	$31,028 (CAD)	0.60	$51,324 (CAD)
Hancock-Howard7	2013	TF TAVR vs. MM	Canada	3 years	2009	$15,687 (CAD)	0.49	$32,170 (CAD)
Murphy8	2013	TF TAVR vs. MM	UK	Lifetime	NR	£15,885	0.44	£35,956
Sehatzadeh9	2013	TAVR vs. MM	Canada	Lifetime	NR	$15,233 (CAD)	0.628	$24,257 (CAD)
Simons10	2013	TF TAVR vs. MM	USA	Lifetime	2010	$85,600	0.73	$116,500
Brecker11	2014	TAVR vs. MM (EuroSCORE ≥20%)	UK	5 years	NR	£22,009	1.24	£17,718
		TAVR vs. MM (EuroSCORE <20%)	UK	5 years	NR	£21,038	1.51	£13,943
Freeman12	2016	TAVR vs. MM	UK	5 years	2012	£13,655	1.29	£10,533
High Risk
Gada1	2012	TA TAVR vs. SAVR	USA	Lifetime	2012	$100	−0.04	TAVR dominated by SAVR
Gada2	2012	TF TAVR vs. SAVR	USA	Lifetime	2011	$3,164	0.06	$52,733
Neyt3	2012	TF or TA TAVR vs. SAVR	Belgium	1 year	NR	€20,397	0.03	€750,000
Reynolds13	2012	TF TAVR vs. SAVR	USA	12 months	2010	−$1,250	0.068	TAVR dominant
		TA TAVR vs. SAVR	USA	12 months	2010	$9,906	−0.07	TAVR dominated
Doble6	2013	TF or TA TAVR vs. SAVR	Canada	20 years	2010	$11,153 (CAD)	−0.102	TAVR Dominated by SAVR
Fairbairn14	2013	TAVR vs. SAVR	UK	10 years	NR	−£1,350	0.063	TAVR dominates SAVR
Sehatzadeh9	2013	TAVR vs. SAVR	Canada	Lifetime	NR	−$4,642 (CAD)	−0.069	$66,985 (CAD)
Reynolds15	2016	TAVR vs. SAVR	USA	Lifetime	2013	$17,849	0.32	$55,090
Intermediate Risk
Ribera16	2015	Edwards SAPIEN TF TAVR vs. SAVR	Spain	1 year	2012	€8,800	0.036	€148,525
		MedtronicCoreValve TF TAVR vs. SAVR	Spain	1 year	2012	€ 9,729	−0.011	TAVR dominated by SAVR
Moore17	2016	Edwards INTUITY Elite MIS-RDAVR vs. MISAVR vs. SAVR	USA	Lifetime	2016	MIS-RDAVR vs. MISAVR: $4,560; MISAVR vs. SAVR: −$7,181	MIS-RDAVR vs. MISAVR: 0.2; MISAVR vs. SAVR: 0.066	MISAVR dominates SAVR; MIS-RDAVR vs. MISAVR: $22,903

^*ΔEffectiveness and ICERs were calculated using QALYs unless specified to be life-years.

TA, transapical; TAVR, transcatheter aortic valve replacement; MM, medical management; TF, transfemoral; SAVR, surgical aortic valve replacement; MIS-RDAVR, minimally invasive surgical rapid deployment aortic valve replacement; MISAVR, minimally invasive surgical aortic valve replacement; NR, not reported; CAD, Canadian dollars.

TABLE E2.

Summary of cost-effectiveness analysis findings for mitral regurgitation

Author	Year	Population	Comparison	Country	Horizon	Cost year	ΔCosts	ΔEffectiveness*	ICER*
Mealing18	2013	Functional or degenerative, moderate/severe MR with HF and high surgical risk	MitraClip vs. MM	UK	Lifetime	2011	£30,192	2.04	£14,800
Cameron19	2014	Functional or degenerative, moderate/severe MR with HF and high surgical risk	MitraClip vs. MM	Canada	Lifetime	2013	$40,617 (CAD)	1.73	$23,433 (CAD)
Armeni20	2016	Functional,moderate/severe MR with HF	MitraClip vs. MM	Italy	Lifetime	NR	€23,342	3.01	€7,908
Asgar21	2016	Functional, moderate/severe MR with HF and high surgical risk	MitraClip vs. MM	Canada	10 years	2013	$52,600 (CAD)	1.63	$32,300 (CAD)
Guerin22	2016	Functional or degenerative, moderate/severe MR with HF and high surgical risk	MitraClip vs. MM	France	5 years	2011	€26,974	1.71 (LY)	€15,741 (/LY)

^*ΔEffectiveness and ICERs were calculated using QALYs unless specified to be life-years.

MR, mitral regurgitation; HF, heart failure; MM, medical management; CAD, Canadian dollars; NR, not reported; LY, life-year.

TABLE E3.

Summary of cost-effectiveness analysis findings for atrial fibrillation

Author	Year	Population	Comparison	Country	Horizon	Cost year	ΔCosts	ΔEffectiveness*	ICER*
Lamotte23	2007	Coronary or valvular disease undergoing CABG or valve replacement/repair with permanent AFib	High-intensity focused ultrasound surgical ablation vs. maze vs. percutaneous ablation vs. MM	UK	5 years	2005	Percutaneous ablation vs. surgical ablation: £971; surgical ablation vs. maze: £1,334; maze vs. no ablation: £720	Percutaneous ablation vs. surgical ablation: −0.083; surgical ablation vs. maze: −0.0233; maze vs. no ablation: 0.536	Percutaneous ablation dominated by surgical ablation; surgical ablation dominated by maze; maze vs. no ablation: £1,343
		Coronary or valvular disease undergoing CABG or valve replacement/repair with persistent AFib	High-intensity focused ultrasound surgical ablation vs. maze vs. percutaneous ablation vs. MM	UK	5 years	2005	Percutaneous ablation vs. surgical ablation: £1,010; surgical ablation vs. maze: £1,284; maze vs. no ablation: £885	Percutaneous ablation vs. surgical ablation: −0.1082; surgical ablation vs. maze: 0.0362; maze vs. no ablation: 0.255	Percutaneous ablation dominated by surgical ablation; surgical ablation vs. maze: £35,469; maze vs. no ablation: £3,471
		Coronary or valvular disease undergoing CABG or valve replacement/repair with paroxysmal AFib	High-intensity focused ultrasound surgical ablation vs. maze vs. percutaneous ablation vs. MM	UK	5 years	2005	Percutaneous ablation vs. surgical ablation: £981; surgical ablation vs. maze: £1,284; maze vs. no ablation: £856	Percutaneous ablation vs. surgical ablation: −0.077; surgical ablation vs. maze: 0.0352; maze vs. no ablation: 0.286	Percutaneous ablation dominated by surgical ablation; surgical ablation vs. maze: £36,477; maze vs. no ablation: £2,991
Quenneville24	2009	Chronic AFib undergoing MV surgery	Concomitant modified Maze vs. MM	Canada	15 years	NR	$900 (CAD)	0.20	$4,446 (CAD)
van Breugel25	2011	Paroxysmal, persistent, or permanent AFib undergoing valvular and/or coronary surgery	Concomitant ablation surgery vs. MM	Netherlands	1 year	2004	€4,426	0.06	€73,359
Anderson26	2014	Symptomatic non-paroxysmal AFib - low event rate risk cohort	Convergent procedure vs. catheter ablation vs. MM	USA	5 years	2013	Convergent vs. Catheter ablation: −$357; Catheter ablation vs. MM: $15,809; Convergent vs. MM: $15,452	Convergent vs. Catheter ablation: 0.23; Catheter ablation vs. MM: 0.52; Convergent vs. MM: 0.75	Convergent dominates catheter ablation; Convergent vs. MM: $20,640
		Symptomatic non-paroxysmal AFib - medium event rate risk cohort.	Convergent procedure vs. catheter ablation vs. MM	USA	5 years	2013	Convergent vs. Catheter ablation: −$4,475; Catheter ablation vs. MM: $6,300; Convergent vs. MM: $1,825	Convergent vs. Catheter ablation: 0.26; Catheter ablation vs. MM: 0.56; Convergent vs. MM: 0.82	Convergent dominates catheter ablation; Convergent vs. MM: $2,214
		Symptomatic non-paroxysmal AFib - high event rate risk cohort.	Convergent procedure vs. catheter ablation vs. MM	USA	5 years	2013	Convergent vs. Catheter ablation: −$8,337; Catheter ablation vs. MM: −$6,336; Convergent vs. MM: −$14,673	Convergent vs. Catheter ablation: 0.28; Catheter ablation vs. MM: 0.62; Convergent vs. MM: 0.90	Convergent dominates catheter ablation; Convergent dominates MM

^*ΔEffectiveness and ICERs were calculated using QALYs unless specified to be life-years.

CABG, coronary artery bypass graft; AFib, atrial fibrillation; MM, medical management; MV, mitral valve; CAD, Canadian dollars; NR, not reported.

TABLE E4.

Summary of cost-effectiveness analysis findings for coronary artery disease

Author	Year	Population	Comparison	Country	Horizon	Cost year	ΔCosts	ΔEffectiveness*	ICER*
Eefting27	2003	Single or multivessel	Off-pump CABG vs. PCI stents	Netherlands	1 year	1999	7,332.5 Dutch Florins	−0.03	Off-pump CABG dominated by PCI
Hlatky28	2004	Multivessel	CABG vs. PCI no stents	USA	12 years	2002	$2,250	0.16 (LY)	$14,300 (/LY)
Nathoe29	2005	Single or multivessel	Off-pump CABG vs. PCI stents	Netherlands	1 year	1999	€2,813	−0.03	Off-pump CABG dominated by PCI.
Stroupe30	2006	Medically refractory myocardial ischemia, high risk adverse outcomes	CABG vs. PCI (some stenting); urgent revascularization	USA	5 years	2004	$18,732	−0.05	CABG dominated by PCI
Kastanioti31	2007	Single or multivessel	CABG vs. PCI vs. MM	Greece	1 year	NR	−3447	0.03	NR
Magnuson32	2013	Multivessel w/ diabetes	CABG vs. PCI DES	USA	Lifetime	2010	$5,392	0.663	$8,132
Cohen33	2014	Multivessel or left main	CABG vs. PCI DES	USA	Lifetime	2010	$5,081	0.307	$16,537
Javanbakht34	2014	Multivessel	CABG vs. PCI stents	Iran	Lifetime	2011	−$4,761	0.41	CABG dominated PCI
Zhang35	2015	Multivessel	CABG vs. PCI; non-emergent	USA	Lifetime	NR	$11,575	0.38	$30,454
Yock36	2003	Multivessel	CABG w/ provisional stent in follow-up PCI vs. Initial PCI w/ provisional stent vs. CABG w/o stent in follow-up PCI vs. CABG w/ primary stent in follow-up PCI vs. initial PCI w/ primary stent	USA	Lifetime	2000	Initial PCI w/ primary stent vs. CABG w/ primary stent in follow-up PCI: $3,800; CABG w/ primary stent in follow-up PCI vs. CABG w/o stent in follow-up PCI: $4,100; CABG w/o stent in follow-up PCI vs. initial PCI w/ provisional stent: $200; Initial PCI w/ provisional stent vs. CABG w/ provisional stent in follow-up PCI: $300	Initial PCI w/ primary stent vs. CABG w/ primary stent in follow-up PCI: −0.32; CABG w/ primary stent in follow-up PCI vs. CABG w/o stent in follow-up PCI: 0.02; CABG w/o stent in follow-up PCI vs. initial PCI w/ provisional stent: 0.34; Initial PCI w/ provisional stent vs. CABG w/ provisional stent in follow-up PCI: −0.35	Initial PCI w/ primary stent dominated by CABG w/ primary stent in follow-up PCI; CABG w/ primary stent in follow-up PCI vs. CABG w/o stent in follow-up PCI: $205,000; CABG w/o stent in follow-up PCI vs. initial PCI w/ provisional stent: $588.24; Initial PCI w/ provisional stent dominated by CABG w/ provisional stent in follow-up PCI
Griffin37	2007	Rated appropriate for CABG	CABG vs. PCI vs. MM	UK	6 years	2003/2004	CABG vs. PCI: £3,230; PCI vs. MM: £2,640	CABG vs. PCI: 0.15; PCI vs. MM: 0.25	CABG vs. PCI: £22,000; PCI vs. MM: £11,000
		Rated appropriate for PCI	CABG vs. PCI vs. MM	UK	6 years	2003/2004	CABG vs. PCI: £4,947; PCI vs. MM: £2,847	CABG vs. PCI: −0.07; PCI vs. MM: 0.06	CABG dominated by PCI; PCI vs. MM: £47,000
		Rated appropriate for CABG and PCI	CABG vs. PCI vs. MM	UK	6 years	2003/2004	CABG vs. PCI: £3,820; PCI vs. MM: £3,435	CABG vs. PCI: 0.24; PCI vs. MM: 0.15	CABG vs. MM: £19,000; PCI extendedly dominated
Eisenstein38	2009	Two vessel with normal-mild CKD	CABG vs. PCI vs. MM	USA	3 years	NR	CABG vs. PCI: $3,079; CABG vs. MM: $9,999; PCI vs. MM: $6,891	CABG vs. PCI: −0.050; CABG vs. MM: 0.038; PCI vs. MM: 0.058 (LY)	CABG vs. PCI: CABG dominated by PCI; CABG vs. MM: $332,506; PCI vs. MM: $140,129 (/LY)
		Three vessel with normal-mild CKD	CABG vs. PCI vs. MM	USA	3 years	NR	CABG vs. PCI: −$1,561; CABG vs. MM: $5,363; PCI vs. MM: $5,593	CABG vs. PCI: 0.098; CABG vs. MM: 0.329; PCI vs. MM: 0.162 (LY)	CABG vs. PCI: CABG dominates PCI; CABG vs. MM: $20,299; PCI vs. MM: $38,582 (/LY)
		Left main with normal-mild CKD	CABG vs. PCI vs. MM	USA	3 years	NR	CABG vs. MM: $15,491	CABG vs. MM: 0.599 (LY)	$28,588 (/LY)
		Two vessel with moderate-severe CKD	CABG vs. PCI vs. MM	USA	3 years	NR	CABG vs. PCI: $8,375; CABG vs. MM: $4,482; PCI vs. MM: −$3,375	CABG vs. PCI: 0.251; CABG vs. MM: 0.360; PCI vs. MM: 0.067 (LY)	CABG vs. PCI: $36,593; CABG vs. MM: $15,661; PCI vs. MM: PCI dominates MM (/LY)
		Three vessel with moderate-severe CKD	CABG vs. PCI vs. MM	USA	3 years	NR	CABG vs. PCI: $20,370; CABG vs. MM: $23,264; PCI vs. MM: −$787	CABG vs. PCI: 0.407; CABG vs. MM: 0.274; PCI vs. MM: 0.003 (LY)	CABG vs. PCI: $54,902; CABG vs. MM: $91,583; PCI vs. MM: $89,364 (/LY)
		Left main with moderate-severe CKD	CABG vs. PCI vs. MM	USA	3 years	NR	CABG vs. MM: $549	CABG vs. MM: 0.729 (LY)	$3,709 (/LY)
Nathoe29	2005	Single or multivessel	Off-pump vs. on-pump CABG	Netherlands	1 year	1999	−€2,089	−0.01	€ 208,900
Shiga39	2007	Single or multivessel	Elective off-pump vs. on-pump CABG	USA	Lifetime	2005	−403	0.12	Off-pump dominated on-pump
Al-Ruzzeh40	2008	Eligible for isolated CABG	Off-pump vs. on-pump CABG	UK	6 months	2003/2004	−£1,478	−0.007	£211,142.85
Houlind41	2013	Single or multivessel	Off-pump vs. on-pump CABG	Denmark	6 months	2010	−10,927.9 D.Kr	−0.0016	6,829,999 D.Kr
Wagner42	2013	Elective or urgent CABG-only surgery	Off-pump vs. on-pump CABG	USA	1 year	2010	$3,601	−0.31	Off-pump dominated by on-pump
Reeves43	2004	Single vessel LAD	MIDCAB vs. PCI stents and no stents	UK	1 year	NR	$892	0.02	$44,600
Rao44	2007	Single vessel LAD	MIDCAB vs. PCI w/ stents	UK	10 years	NR	$829.02	0.132	$6,274.02
Rao45	2008	CABG patients	Minimally invasive vs. conventional vein harvesting in CABG	UK	6 weeks	NR	$458.74	0.0231	$19,858.87
Oddershede46	2012	CABG patients	Endoscopic vs. open vein harvest in CABG	Denmark	35 days	2011	$216.74	0.00273	$79,391
Hatky47	2009	Stable CAD, diabetes 2	Prompt CABG vs. MM w/ delayed CABG	USA	Lifetime	2007	$25,000	0.494	$50,000

^*ΔEffectiveness and ICERs were calculated using QALYs unless specified to be life-years.

CABG, coronary artery bypass graft; PCI, percutaneous coronary intervention; LY, life-year; NR, not reported; MM, medical management; LAD, left anterior descending artery; MIDCAB, minimally invasive direct coronary artery bypass; CAD, coronary artery disease; CKD, chronic kidney disease.

TABLE E5.

Essential criteria of a useful model-based cost-effectiveness analysis

Criterion*	Requirements
Relevant decision problem	The model addresses a medical decision problem with clear trade-offs between the potential benefits and harms among the considered interventions
Representative patient cohort	The model simulates a cohort representative of the target patient population or individual (e.g. “average”) patient
Exhaustive comparisons	All relevant, competing intervention strategies that can be considered for the decision problem are included in the model
Appropriate outcomes	The cost and quality-of-life estimates are applicable to the (envisioned) clinical practice and analytic perspective (e.g. healthcare sector / societal), also the modeled fatal and non-fatal event rates are applicable to the target population/patient
Appropriate time horizon	The model is capable to make projections over a sufficiently long time horizon in order to capture all relevant future costs, beneficial and harmful health outcomes
Transparent model	The model and input parameters are well described, assumptions are clear and valid, and results on both intermediate (e.g. event rates, cumulative costs per event type) as primary model outcomes (aggregated cost and effectiveness outcomes) are presented
Credible and plausible model output	The model generates output that matches with what can be expected from the current knowledge and plausible explanations are provided if that is not the case
Internally valid and generalizable predictions	Predictions by the model are in agreement with observations from the underlying data source(s) and, especially when the underlying data have limited sample size or have been modified, also with observations from external independent data
Experienced research team	The research team is experienced and team members have a track record of published cost-effectiveness analyses

^*Partly adapted from Habbema et al.⁴⁸

FIGURE E1. Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram of study inclusion.

Central Picture. Publication trends in cardiac surgery cost-effectiveness analysis since January 2000.

CENTRAL MESSAGE.

Cost-effectiveness, a measure of economic value increasingly applied to cardiac surgical procedures, is essential for the rational adoption of new interventions given health care budget constraints.

PERSPECTIVE STATEMENT.

Cost-effectiveness analysis in cardiac surgery continues to grow in relevance with an increasing emphasis on value-based care and the expansion of high cost devices and procedures. Economic data are increasingly being gathered within clinical trials and in cardiac surgery registries, providing opportunities to integrate economic outcomes into an evolving surgical practice.

GLOSSARY OF ABBREVIATIONS

AF

atrial fibrillation

AS

aortic stenosis

BTT

bridge to transplant

CABG

coronary artery bypass graft

CAD

coronary artery disease

CEA

cost-effectiveness analysis

DES

drug eluting stent

DRG

diagnosis related group

DT

destination therapy

HF

heart failure

HRQoL

health-related quality-of-life

ICER

incremental cost-effectiveness ratio

LAD

left anterior descending artery

LVAD

left ventricular assist device

LY

life year

MIDCAB

minimally invasive direct coronary artery bypass

MR

mitral regurgitation

PCI

percutaneous coronary intervention

PSA

probabilistic sensitivity analysis

QALY

quality-adjusted life year

QoL

quality-of-life

RCT

randomized controlled trial

SAVR

surgical aortic valve replacement

TAVR

transcatheter aortic valve replacement

US

United States