Background

Recent emphasis on cost-containment in the health care environment has focused attention on the cost of medical procedures. Selection of the appropriate treatment for coronary artery disease is of increasing concern. Coronary artery bypass grafting is common and very expensive, and this procedure will continue to be examined closely by reimbursement systems, particularly with regard to the lower initial cost of coronary angioplasty as a competing therapy.

Methods

Duke University Medical Center has a sophisticated accounting system that enables individual cost components to be identified, facilitates prospective analysis of cost/benefit, and aids allocation of limited hospital resources. In 1996, 1,114 coronary artery bypass procedures were performed at Duke. Preoperative patient characteristics were also analyzed in an attempt to predict risk factors for increased cost.

Results

The median cost for these procedures was $20,682, excluding professional fees. Sixty percent of the costs were directly associated with patient care, and the other 40 percent were accounted for by indirect costs to support patient care. The most significant preoperative predictor of increased postoperative cost was the mortality estimate. If this variable was excluded from the analysis, other variables (for example, ejection fraction, age, identity of the surgeon, and congestive heart failure) were all related to increased costs.

Conclusion

Predicting costs based on preoperative variables offers the potential to reduce total costs through case-management strategies and aids in negotiating a risk-shared contract. However, cost reduction in routine care will have more financial impact than cost reduction by patient selection.

Cost prediction in coronary artery bypass grafting (CABG) can facilitate prospective comparison of cost and patient benefit, identify unnecessary costs and the relative value of accrued costs, aid prospective resource allocation for managed-care contracting, and maximize the use of limited hospital resources.

This report presents information about cost derived from a data set at Duke University. Duke has a sophisticated cost-accounting system that provides substantial data about how we are faring economically with our patients. It should be kept in mind that true cost varies by institution, making comparisons problematic. Also, total cost includes therapy by providers at other institutions, and data on costs accrued outside of Duke are not available. Total cost also should include the opportunity cost for both the patient and members of the patient’s family, who become the patient’s caregivers after a shortened hospital stay. Indirect costs usually are influenced by joint physician action, and only rarely by the individual surgeon’s decisions.

Cost of care may be considered to include necessary and unnecessary determinants. I arbitrarily define necessary cost determinants as the factors associated with routine care and the management of complications that occur with some regularity, although without much predictability. I define unnecessary cost determinants as protective practices undertaken to defend against legal liability and anecdotal or non-data-driven practice patterns.

The analyses presented below are of data derived from 1,114 CABG procedures performed at Duke University Medical Center between December 1995 and December 1996. Mean patient age was 65 years; mean ejection fraction was 0.55. Six percent of the procedures were reoperations. Thirty-four percent of the patients were female, compared with only 28 percent four years earlier. Thirty-four percent of the patients had diabetes, taking either oral agents or insulin. The median cost for these procedures was $20,682, excluding professional fees.

Variable and Fixed Direct Costs; Indirect Cost

Duke’s accounting system enables costs to be broken down into variable direct cost, fixed direct cost, and indirect cost. Variable direct cost is the marginal cost of performing the next coronary procedure contracted for. Fixed direct costs are costs directly associated with patient care, which do not vary with incremental volume of care. Indirect costs are costs associated with services that indirectly support patient care.

For these 1,114 CABG procedures, the variable direct cost accounted for only about 50 percent of the total cost. This is the only component of CABG cost that we can affect individually as surgeons. The fixed direct cost accounted for about 10 percent of total cost. The indirect cost was a huge 40 percent of total cost, reflecting the tendency of hospital administrations to preferentially assign indirect costs to the surgical services, because that is where the money is.

Analysis of the departmental cost distributions for CABG without cardiac catheterization shows that 38 percent of the total cost was attributable to the operating room, 12 percent to the intensive care unit, 13 percent to intermediate care, 8 percent to regular room and board, 4 percent to blood requirements, and 6 percent to anesthesia, with the remainder accounted for by cardiology, radiology, the laboratory, and the pharmacy.

When patients were plotted by frequency distribution of cost in increments of $2,000, by far the largest portion of the total cost of coronary operations at Duke in 1996 (somewhat less than $30 million) was accounted for by patients whose total cost fell within the range of $12,000 to $32,000 (Fig. 1).

Preoperative Variables as Predictors of Cost

A multivariate analysis was done of the preoperative variables associated with these patients in an attempt to identify which factors contributed to increased cost. Predicted costs in the model used for this analysis were subsequently found to correlate very well, although not perfectly, with actual costs. The most powerful preoperative predictor of postoperative cost, as indicated by C ² values, was the mortality estimate, using the New York State model. The next most powerful predictor was catheterization done at Duke; because these patients accrued cost from their catheterization, this finding was expected. The identity of the surgeon was a very significant factor determining cost, despite the use of clinical pathways to minimize surgeon variance in practice. Ejection fraction was the fourth most powerful variable, with lower ejection fraction significantly increasing cost, despite the fact that ejection fraction is a component included in the mortality estimate. Congestive heart failure, cardiomegaly, and unstable angina were the only other variables that were significant (Table 1).

When we ran the data again, omitting the mortality estimate from the equation, ejection fraction became the single most powerful predictor of excess cost, followed by patient age and then identity of the surgeon. Congestive heart failure, unstable angina, cardiomegaly, reoperative coronary surgical procedure, and diabetes were all significant factors related to increased cost.

Univariate analysis of the effect of predicted mortality on cost in this model showed a predicted mortality of 1 percent associated with a cost of about $20,000 and a predicted mortality of 17 percent associated with a cost of about $35,000. Univariate analysis of the effect of age showed a general trend toward older groups having higher costs; however, the univariate effect was not consistent, because patients had varying degrees of diabetes, congestive heart failure, coronary disease, and other factors.

Cost variance by surgeon in our group, with all surgeons doing more than 100 CABG procedures for the year, ranged from $20,000 to about $30,000 (Fig 2). The actual mortality rates of all surgeons were below their predicted mortality rates, and there was no relationship between actual mortality and cost.

Univariate analysis showed that the presence or absence of certain preoperative comorbidities affected cost outcome (Table 2). Our cost data also showed a survivorship effect. The average cost per patient for 22 non-survivors among 1,074 patients was $60,320; this compared with an average cost of $23,826 for surviving patients (Table 3). Some costs for non-survivors were very low and others were extremely high, as would be expected.

Predictors of early discharge were also identified in this data set. Factors found to be significantly associated with early discharge were lower age, high ejection fraction, absence of diabetes, male sex, and no prior percutaneous transluminal coronary angioplasty to early discharge remains unexplained.

Effect of Patient Selection by Predicted Mortality Stratification

Stratification of patients by predicted mortality showed that the average cost per patient among 23 patients with predicted mortality greater than 15 percent was $33,049; this compared with an average cost of $24,388 among patients with predicted mortality of less than 15 percent (Table 5). But a look at system costs showed that the denial of operation to patients with greater than 15 percent predicted mortality would have saved only $760,132 out of a total system cost of more than $26 million, thus saving only about 2.3 percent of total cost while denying the benefit of operation to 2.1 percent of patients. These data suggest that the selection of patients on the basis of risk is not the way to go. Coronary surgery is characterized by high-risk; high-cost, and high-patient benefit.

Summary

Preoperative variables can be used to predict costs. This provides the ability to estimate the cost of caring for a given patient population, important in negotiating a risk-shared contract. It also offers the potential to reduce the cost of caring for particular patients through case management strategies, thereby reducing total system costs.

Surgeon practice can significantly affect cost without adversely affecting outcome. If we do not address this factor, others will. At some point, there will be "economic credentialing" of surgeons who cannot provide cost-effective practice. Managed-care companies may begin to discriminate against surgeons who cannot achieve the same cost-effectiveness as other surgeons while achieving similar outcomes.

Cost reduction in routine care has more impact, and more appropriate impact, than cost reduction by patient selection. Economic stratification is a dangerous tool. It involves considerations that will require political solutions rather than solutions by individual surgeons.

Table 1. Preoperative Variables Associated With Cost
 

Table 2. Preoperative Comorbidity
 

Table 3. Survivorship Effect
 

Table 4. Predictors of Early Discharge
 

Table 5. Predicted Mortality Stratification