Medical Economics in Interventional Cardiology

Evaluation of Procedure and Hospital Cost

Evaluation of the true cost of a specific procedure and the associated hospital stay is a very challenging process. In practice, obtaining the detailed data about individual resources being consumed for an intervention and adding up these costs to perform marginal or incremental cost analysis (the bottom-up approach) are difficult and often impractical. Performing a true bottom-up cost analysis (microcosting analysis) is a complex, time-consuming process that requires identification of all the inputs into a health care service and the assignment of an appropriate cost to each. Furthermore, even if this process was achievable for simple treatments, it becomes more complex when considering complex interventional procedures such as a PCI and very complicated when accounting for the entire hospital stay from admission to discharge. Therefore most U.S. cost studies start with an aggregated measure of costs, such as can be obtained from hospital or physician bills (a top-down analysis).

To overcome this complexity, cost data in the United States are obtained mainly by one of two approaches: The first approach involves converting hospital charges (taken from the hospital bill) to costs using the ratios of costs to charges (RCCs) included in each hospital’s annual Medicare Cost Report. This report is built based on annual reporting of every hospital to the Centers for Medicare and Medicaid Services (CMS) that includes details of expenses for direct patient care, overhead, capital equipment, and so forth related to billed charges. Although the Medicare RCCs are not designed for research, it still provides a moderately standardized means of estimating cost across hospitals in the United States. In addition, costs calculated with the RCC method are used to recalibrate diagnosis-related group (DRG) weights by CMS. Therefore this method provides a valuable tool for multicenter cost research. However, this approach has several limitations. First, this approach does not separate out overhead and most other fixed costs and therefore provides only an estimate of average rather than marginal cost, so it may overestimate potential cost savings. Second, due to the complexity of the instructions for reporting to CMS, hospitals may choose to interpret the instructions differently, resulting in significant variability in uncovering hospitals’ actual costs. Finally, the RCCs themselves are an average of all cost-charge relationships from hospital revenue centers, such as the radiology, pharmacy, or laboratory departments; therefore the Medicare whole hospital-level RCCs may not be particularly accurate in converting these charges to costs to reflect individual patient resource consumption. In addition, this method is limited to hospitals using standardized methods of billing, applicable to most but not all U.S. hospitals and to no hospitals outside the United States. Finally, physician costs are not included in this approach.

The second approach depends on calculating the hospital and intervention costs from reimbursement rates for DRGs reported to the CMS by hospitals. This approach also has several limitations. First, it is not sensitive to variations in resource-use intensity within a DRG, as DRG reimbursement rates represent the “average” cost for a particular diagnosis/procedure among all patients in that DRG. Second, the CMS decides in many cases not to increase reimbursement to cover the costs of new technology; therefore the cost of these new technologies may not be well reflected in the DRG. However, this approach does not require collection of hospital bills and, with some assumptions, can potentially be applied to any hospitalization in any health care system, therefore simplifying its use. By assigning DRGs to hospitalizations within studies where DRGs from the hospital bill is not available, this approach can be used for international studies. However, assigning costs to a DRG for hospitalization in international studies is complicated and controversial.

Many studies estimate costs by counting only big-ticket items used (e.g., number of diagnostic cardiac catheterizations, PCIs, or coronary artery bypass graftings (CABGs); days in the intensive care unit [ICU]; total hospital length of stay) and assigning a unit price to each item. The resulting linear formula: Total cost = Σ price × quantity is simple and inexpensive to use (which makes it desirable in clinical research); however, this approach suffers from some significant downsides. First, the source of cost weights is often acquired from available unrelated economic sources external to the resource data being analyzed and therefore of uncertain quality. Second, the appropriate set of big-ticket items necessary to estimate costs accurately by this method has never been rigorously defined. For example, some studies, especially multicenter clinical trials, may use the more easily obtained total hospital length of stay instead of defining specific ICU and non-ICU hospital lengths of stay that may be more representative of accurate cost. Third, most studies treat big-ticket items similarly to preserve the desired simplicity, although they are not necessarily homogeneous. For example, by using payments by the DRG, an uncomplicated single-vessel PCI would typically be assigned the same price as a complex three-vessel PCI or complex graft intervention that was complicated by abrupt closure, while the true costs of these procedures are, in fact, substantially different.

Evaluation of Physician Costs

Evaluating physician costs also has serious challenges. This estimation is also made by one of two approaches. The first approach depends on acquiring the “fee for service” bills from physicians’ offices. This approach had several limitations. First, most patients receive services from several physicians and collecting these bills from multiple offices is a far more complicated process than collecting hospital bills. Second, due to multiple services with different costs and reimbursements in physicians’ offices, cost-shifting process is used to cover for unreimbursed and underreimbursed services without the detailed report of real costs that is required by hospitals when they practice cost shifting. Therefore using “fee for service” bill is not an accurate reflection of the true market price for physician services.

To overcome the limitation of using the physicians’ bill, another approach uses the Medicare Fee Schedule based on the resource-based relative-value scale (RBRVS). An estimation of physician costs can be created using available data from the American Medical Association Physician’s Current Procedural Terminology (CPT). This approach is not ideal, but it has the advantage of being more objective and consistent and might represent the best available U.S. national measure of the economic value of physician work.

Factors That Impact Cost Assessment

The different perspectives of patients, providers, health payers, and society will lead to different assessment of cost. Although patients may consider only out of pocket or copayment costs, health payers may consider the cost of the procedure, associated hospital cost, and whether the medication used is generic or nongeneric. Providers may consider cost in relation to the health benefit for an individual patient and then weigh it against public health implications of this procedure. When assessing the cost of a health service, ideally it should assess not only the immediate cost of the strategy but also consider the overall society perspective, which ultimately rests with all stakeholders in society. Societal costs include the hospital costs, physician fees, outpatient testing, and outpatient drug therapy costs, but also nonmedical direct expenses (e.g., transportation to the medical facility, child care, housekeeping) and the economic impact of lost productivity because of illness.

However, total societal cost cannot be measured effectively because each component is measured by a proxy. To eliminate some of this uncertainty, payer costs are often measured instead. However, payer costs do not represent the complete spectrum of the societal costs. CMS uses DRGs to pay hospitals regardless of the cost of service to the provider, whereas private insurance may use a different scheme to pay providers. Payments may therefore have significant variability between payers.

Furthermore, time effects are important to medical cost analyses because either inflation or deflation will change the cost of a procedure. A common approach to deal with this is to pick a base year and then use the medical inflation rate to deflate all costs in subsequent year to the base year.

Another important factor in determining cost is related to geography. The costs of material and labor inputs to medical care can vary substantially from one part of the United States to another and between large urban university hospitals to small rural community hospitals, creating true differences in cost for medical services according to geography that needs to be adjusted. Several geographic adjustment indices are available, including the Medicare area wage index (for adjusting DRG reimbursement) and the Medicare Fee Schedule geographic adjustment factor. International studies are yet more complex.

Utility Assessment

Utility may be measured by patient preference methods or by surveys that have been mapped to patient preference methods. Theoretically, the best direct measure of utility is the “standard gamble” method because it assesses both overall quality of life and risk aversion. However, this measure is complex to obtain and may not be practical to administer in large trials. Most economic analyses derive utilities from surveys (e.g., EQ-5D or the Health Utilities Index). Using this approach, patient health states are defined by a number of explicit domains such as physical functioning and pain. Then, previously measured utility weights from patients (i.e., those with the condition of interest) or the general public (i.e., those who are at risk for getting the condition of interest) are assigned to each possible health state.

Sensitivity Analysis

Even in the most carefully conducted cost-effectiveness studies, considerable uncertainty may remain regarding the parameters used to measure costs and health effects. This uncertainty is in addition to accounting for the play of chance with bootstrap analysis, which is not sufficient to account for any additional uncertainty or bias. To identify and assess the degree to which uncertainty in the parameter could affect the overall results, CEAs usually perform multiple evaluations in which one or more of the parameters are varied across reasonable ranges. The ranges reflect intrinsic variability or regional variation. As an example, the cost of diagnostic cardiac catheterization may be cheaper in some institutions or health care delivery settings compared with others; thus to determine the cost effectiveness of PCI in stable coronary artery disease (CAD), the cost of PCI could be varied over a range to determine the maximal cost at which it remains cost effective. This process (termed “sensitivity analysis”) allows for a reasonable appraisal about the parameters that are most important in the analysis and the stability of the reference case results.

Sensitivity analyses have typically assessed the effect varying selected parameters related to cost effectiveness one at a time. A more contemporary approach called Monte Carlo simulation permits all parameters to be varied simultaneously. These sophisticated analyses yield a cost-effectiveness acceptability curve, which accounts for uncertainty in all model estimates. The end result is that the curve displays the likelihood that a new intervention will have a cost-effectiveness ratio that falls below a particular societal “willingness to pay.” (See Benchmarks in Economic Analysis later.)