The relationship between hospital charges and a modified parsonnet risk source.

Economic planning is of paramount importance in hospital resource allocation. At Evanston Hospital, we have attempted to use the Parsonnet score as a preoperative discreet independent variable to predict hospital charges in CABG patients. Our objective was to develop a simple formula to predict hospital charges in a prospective manner, given any patient cohort.

The patient cohort includes all patients undergoing CABG. The group includes all redo procedures and emergency procedures following angioplasty failures but excludes patients in whom any additional cardiac procedure, such as a valve replacement/ repair or aneurysm resection, were done. We analyzed 219 consecutive patients in DRG 106 and 107 operated upon at the Evanston Hospital during the 1991 fiscal year. The clinical demographic features of the group are shown in table 1, page 33. A modified Parsonnet score was determined for each patient (table 2, page 33). Descriptive age, charge, and Parsonnet score statistics are shown in table 3, page 34. Each patient's total hospital charges were recorded and compared to his or her Parsonnet score.

Table 1. Clinical Demographic Features

Males              78%
Females            22%
LIMA               78%
Redo                6%
IABP (preop)       11%
Angina Class IV    63%
Mortality         2.4%

Table 2. Modified Parsonnet Score
                                   Assigned Weight

Female                                    1
Obesity                                   3
Diabetes                                  3
Hypertension                              3
Ejection Fraction Score
    Good >50%                             0
    Fair 30-49%                           2
    Poor <30%                             4
Age Score
    0-74                                  7
    75-79                                 12
    >80                                   20
Reoperation
    First                                 5
    Second                                10
Preoperative IABP                         2
LV Aneurysm                               5
Emergency Surgery Following PCTA
   or Cath Lab Complications              10

Total Score                               Sum

[TABULAR DATA OMITTED]

Regression analysis using the Parsonnet score as the independent variable was done to determine its relationship to total hospital charges. Using uni-variant regression analysis, each element of the Parsonnet score, as well as the total Parsonnet score, was evaluated. Results of this analysis are recorded in table 4, page 34, which shows the correlation coefficient. Multiple regression analysis was also done using combinations of the Parsonnet elements and is shown in table 4.

Results

Most of the elements of the Parsonnet score correlated in a positive way with total charges. However, by themselves only reoperation and ejection fraction score had correlation coefficients equal to or above 0.30. From a practical standpoint, reoperation represents a small subset of this group and, as such, is not a useful predictor of total hospital charges for the group as a whole. Ejection fraction score, on the other hand, is useful as a predictor of total hospital charges. When the total Parsonnet score is compared to charges, a correlation coefficient of 0.30 is obtained. The relationship between total charges and the total Parsonnet score is represented graphically for the group as a whole and for each DRGs (figures 1, 2, and 3) The multivariant regression relationships, which include reoperative score and ejection fraction score, while having stronger correlation coefficients, are a weak predictor for the group as a whole because of the small subset of reoperations. The relationship of increased charges associated with patients undergoing reoperation, however, for that subgroup is an important relationship because more reoperative coronary surgery is being done.

When the total Parsonnet score is used to develop a predictive formula for hospital charges, the "Y" intercept point for DRG 106 falls at $33,100 and for DRG 107 at $25,000. The coefficient for the Parsonnet score is $754 for DRG 106 and $507 for DRG 107. The predictive formulas are shown below.

Discussion

We have demonstrated a positive correlation between the preoperative modified Parsonnet score and total hospital charges. The formulas generated from this relationship enable us to predict total hospital charges on the basis of a prospective analysis of surgical mortality/morbidity risk factors. The predictive formulas enable us to calculate total hospital charges for a cohort of patients for whom we know the Parsonnet score. Reimbursement methodology currently does not take into consideration stratification characteristics such as the Parsonnet elements we have used. An individual provider, however, could examine the cohort of patients it is currently treating and predict whether or not it will be profitable in an overall sense to provide care to the group.

The ability to predict resource consumption necessary to accomplish a treatment protocol, in this case coronary artery bypass grafting, when compared to the revenue obtained for providing the service enables the profitability of the enterprise to be determined.

A variety of interventions can be undertaken once the profit relationship is understood. Efforts can be focused to reduce the resource consumption necessary to provide the care. Changes in reimbursement methodology based on risk stratification can be implemented in order to increase reimbursement in selected cases. An individual institution may elect to discontinue a product line completely as a result of understanding the revenue/cost relationship. It has been customary for hospitals to shift costs to high-volume, high-revenue product lines to finance other product lines that are less profitable. In the future, purchasers of health care are not going to participate in these practices. It will become mandatory for each hospital to very clearly understand and defend its cost structure.

This study was based on hospital charges that, while reliably available, depend on each hospital's unique charge structure and are only an approximation of the actual cost to deliver the service. The hospital accounting practices and assumptions can sometimes cause a distortion in profit/loss analyses. Great care must be exercised when decisions are made for financial reasons. The economic necessity to carefully allocate health care dollars makes understanding the drivers of health care expenditure a fundamental piece of information as resources allocation is undertaken.

This work may serve as a relatively simple but valid model for financial modeling of other DRGs. We have learned a great deal about ourselves from this analysis and believe that we are better positioned to successfully compete in the competitive environment of health care reform.

John C. Alexander Jr., MD, Robert J. Gottner, MD, Carl E. Arentzen, MD, and Robert W. Anderson, MD, are with the Department of Surgery, Evanston Hospital, Northwestern University Medical School, Evanston, Ill.