Efficiency in the field of economics increases when either 1) outputs are increased for a given level of inputs, or 2) inputs are decreased for a given level of output. Estimating efficiency in the medical field is more difficult, however, since the output (marginal health improvement) is difficult to measure. In the area of hospitals, some researchers have estimated the cost per hospital stay, or cost per patient-day as an estimate of efficiency. Newhouse (1994) offers a striking analogy of why this would not be a good measure of efficiency.
“Consider the following analogy. Among the amenities of first-class air travel are wider seats, better food, and a higher ratio of flight attendants to passengers than in the coach section. These amenities are obviously valued because some passengers pay an incremental amount for them. An analog to patient-days or stays in air travel is passenger-miles or the number of passengers; without additional adjustment, however, such output measures would make the additional costs associated with the first class section appear as inefficiency, not as something consumers valued.”
There are other difficulties in measuring hospital efficiency. For instance,
- Many hospital inputs are typically omitted. For instance, capital inputs and the labor costs from physicians with admitting privileges are frequently omitted.
- Case mix controls are not perfect and often hide significant variation in patient illness severity within each diagnostic cost group (DRG).
- When using frontier analysis to measure efficiency, strong assumptions are needed. Data Envelopment Analysis (DEA) assumes no measurement error. [“Random measurement error that is inframarginal (off the frontier) will not affect the location of the frontier, though it will affect how inefficient any given firm appears to be, but sufficiently large error in a given direction will move the frontier itself, thereby increasing the measured inefficiency among firms lying near that segment of the frontier.”] Stochastic frontier estimation allows for measurement error, but makes specific assumptions on the error distribution.
- There are many outputs to estimate. Estimating aggregate efficiency for each hospital does not provide sufficient detail. However, creating an efficiency measure for each disease type or patient types creates a more difficult estimation problem. Further, there is no obviously superior manner in which to aggreate these disease specific scores into a single, hospital-wide efficiency score.
Newhouse J (1994) “Frontier Estimation: How useful a tool for health economics?” Journal of Health Economics, v13(3): 317-322.