A contribution from Prof Pat Rivett
- PDF / 485,444 Bytes
- 1 Pages / 589.56 x 841.92 pts Page_size
- 40 Downloads / 147 Views
The paper by Crofts, Barlow and Edwards, A
market was still more than we could supply. and I put
Clinique Solution, (vol lO, number 4) is of great interest to me personally in my work in the NHS hut
the whole of the OR group for a month on to the problem of producing more large coal. lt never occurred to anyone, including myself, that the set-
also illustrates a general point about modelling which is often overlooked in OR practice.
back for small coal was the first symptom of a future decline in the demand for coal per se.
The classic approach involves listing all the variables
which might he important, separating them into controllable and the uncontrolled, formulating a hypothesis as to the logical interactions of these factors and so on. But we also have to identify the constraints, and the problem is that some of these
This brings us to the Crofts, Barlow and Edwards paper. In this paper the constraint which is accepted is that in their model, the service time, that is the time
the patient spends with the doctor, is treated as a random variable and is averaged out. In practice. this is not the case.
constraints are hidden or too easily accepted as they are frozen out of history.
The doctor is working under severe time constraints, and it is notorious that the first one or two patients in
An example of this is some work on advertising models with which I was associated in the
a session get more than their fair share of time. In some cases I have even found that the consultation
confectionery industry. The company's model work was good and provided estimates of sales in terms of the money spent on advertising. (Indeed it was rather
schedule sheet is printed and sold with 8 minute slots,
so the printer has really set the doctor's schedule.
depressing since although advertising agencies
Consultation time is a constraint and is a controllable variable.
anguish over the theme and content of advertisements, - 'never show a dog with small children as subconsciously mums will be frightened for the kids' - but in the model, all that mattered was
What is really needed is research into the value of the
length of the consultation in terms of the time
the money spent.) On inspection the model related the sales in each four week period to the money spent on advertising in the previous successive four week periods, fitting a discount factor to past data. All very
involved, which is clearly a function of the condition of the patient. Has anyone estimated this value and the knock on effect of that value to the subsequent patient progress? This is of particular concern in the
good. But when one asked why a four week period was chosen we were informed that this was the basic accounting period and all incomes and expenditures were in four week tranches. So the statistical model fitting depended on the convenience of the
paper as it affects the logical underpining the simulation but is far more important than that in patient care generally. Until this is done the
usefulness of simulations of medical appoin
Data Loading...
