|Abstract or Summary
- This thesis develops a cost model for short-term, non-federal,
general hospitals through a regression analysis of data from a sample
of D. S. hospitals. The analysis examines the impact various factors have
on hospital costs as measured by the cost per patient. The results contradict
the beliefs of other researchers and contain some interesting
implications for management of today's hospital system.
Past researchers have been plagued with the problem of accounting
for differences in hospital output. Many believed the most commonly
used measures, patients and patient-days, were inadequate since they do
not account for differences in case mix, complexity of cases, or quality
of the care administered. The possibility of these factors affecting
costs has prompted model builders to devise methods of considering them
in their model.
Two general methods have prevailed. One, epitomized in work done
by Martin Feldstein and Judith and Lester Lave, categorizes patients according to case type and complexity of case. Then a regression model
is developed to explain the differences in hospital costs as measured
by cost per patient or cost per patient day. The best results obtained
was an R²
of .85. However, while being theoretically very appealing,
this approach requires too much inaccessible data to be a pragmatically
The other approach, advanced through work done by Edwards, Miller,
and Schumacher, accounts for output differences in terms of the services
offered instead of those services actually rendered. Through a novel
application of Guttman scaling analysis they developed a scope of service
index to use in their regression model. However, errors in their regression
procedure left the value of the index untested.
The research in this thesis begins by closely examining the work
done by Edwards, Miller, and Schumacher and testing their model in a
regression analysis. The model attained an R² of only .43 so another
method of regression modelling known as the dumpy variable technique
was attempted to see if services could be related more closely to the
dependent variable. This method attained only slightly better results.
Therefore, it appeared that the ability to measure the amount of service
being offered was not the problem with the service adjusted model. Instead,
it indicated that some other important variables were not being
After conducting an exhaustive search for the missing variables,
it was concluded that personnel productivity and average annual wage rate
were the only variables which could adequately explain the differences in hospitals! cost per patient. A model containing these two variables
produced an R² of .88 which seems to show a significant improvement over
models constructed by other researchers.
The overwhelming importance of personnel productivity in explaining
hospital cost performance suggests that for the most part a hospital's
costs are controllable through efficient management of their manpower
resources. But hospitals are in a monopolistic industry and are generally
not motivated by economic factors. Therefore, an external incentive for
controlling costs must be provided.
It is recommended that this incentive be in the form of continued
governmental control of pricing practices.