E. Clinical Cost Function to Find VA Healthcare Cost

1. How do I estimate costs with a clinical cost function?

A clinical cost function can be a useful way to estimate the cost of VA hospital stays. Using non-VA data on hospital stays, one estimates a regression with cost as the dependent variable and characteristics of the stay and the hospital as independent variables. The estimated coefficients from the regression are then applied to VA hospital stays, yielding an estimated cost of each VA stay.

This method requires a suitable source of non-VA data. The dataset must include cost (or cost-adjusted charges) and the factors that are the most influential in explaining the variation in resources, such as the characteristics of the patient, the hospital, and the hospital stay.

Regression-based cost estimates do not capture all variation in hospital expenses. The analyst must decide whether the assumptions of this method are justifiable in the context of a particular study.

Advantages

The cost-regression method requires fewer data elements than is needed to prepare a pseudo-bill. It is difficult to prepare a pseudo-bill for VA inpatient stays because VA doesn't gather the same information that non-VA hospitals and physicians use to bill for their services. An especially important deficiency is information on procedures performed on medical inpatients.

The clinical cost function also has the potential for greater accuracy than an average cost estimate. For some studies, the VA cost analyst does not wish to apply the assumptions used in average costing--for example, the cost of an inpatient stay if proportionate to the Diagnosis Related Group (DRG) weight. There may be additional information available that explains the differences in costs in hospital stays that have the same DRG. Examples of factors that may affect cost are the length of stay, the diagnosis assigned to a hospital stay, the procedures that were performed, or the patient's vital status at discharge.

Physician services

Cost regressions are ordinarily estimated from cost-adjusted charges. In the private sector these exclude physician services because physicians bill separately. VA costs include all physician services, however, and so cost-adjusted charges from non-VA hospitals will understate the true cost of providing services at VA.

How can one account for the cost of physician services? One approach is to assume that physician services are proportional to other costs. The average cost of physician services is expressed as a percentage of the hospital bill and then added to it. Another method is to use the average payment for similar services, derived from published studies. Mitchell et al. (1995) and Miller and Welch (1993) have examined the average Medicare reimbursement for physician services provided to hospitalized patients for each DRG (Diagnosis Related Group). These values, now somewhat dated, must be adjusted to account for inflation and for the greater average length of stay in VA hospital stays relative to Medicare-funded stays.

Statistical considerations

Cost-adjusted charges are ordinarily the dependent variable. Information about the encounter is used as the independent variables. Cost data are not normally distributed. Estimating a function with a skewed (non-normal) dependent variable violates the assumptions of ordinary least squares regressions. This problem can be overcome by transforming the dependent variable, making it more normally distributed. One common method is to take the natural log of cost-adjusted charges. Simulating costs from a function that uses a transformed dependent variable requires correction for retransformation bias.

The analyst will certainly wish to include length of stay as an independent variable in models of the cost of hospital stays. The analysis should avoid the assumption that costs are proportionate to the length of stay, however. It is advisable to include the square and even the cube of length of stay, or to use a specification that allows the parameter to shift as the stay progresses.

Hospital stays at VA facilities are longer than non-VA facilities. It is unlikely that extra days of stay in VA facilities have the same cost as extra days of stay at non-VA facilities. Care must be exercised in simulating VA costs. One approach that has been used is to make the assumption that the median length VA stay has the same cost as the median length non-VA stay, assuming all other factors are held constant. A variation of this method is to estimate a model with the rank of the length of stay, rather that the number of days of stay, as the independent variable.

References

Barnett PG. Determination of VA health care costs. Medical Care Research and Review 2003;60(3 Suppl):124S-141S.

Miller ME, Welch WP. Analysis of Hospital Medical Staff Volume Performance Standards: Technical Report. Washington, DC: Urban Institute, 1993.

Mitchell JB, McCall NT, Burge RT, Boyce S, Dittus R. Heck D, Parchman M, Iezzoni L. Per Case Prospective Payment for Episodes of Hospital Care. NTIS report PB95226023. Waltham, MA: Health Economics Research, 1995.