As many asked for a summary to the following question, here are the
collected responses. thanks for everyone who responded. your collective
advice has been invaluable to us. cheers! arri/justin
We am planning to do an economic evaluation of two alternative diagnostic
tests. Surveying the current literature, many authors report a cost-
effectiveness ratio in the form of 'cost per case diagnosed' for economic
evaluation of tests. However, this approach ignores the economic
consequences of, for instance, false negatives and false postives that
result from the alternative diagnostic tests. Reporting a unitary measure
of effectiveness (cases found) whilst ignoring other substantial outcomes
(value of true negatives, false positives and false negatives) does not
seem to capture the full economic consequences of alternative tests. If
any of you are aware of any literature where this problem has been tackled,
would you please let us know.
Cheers!
Arri
Justin
Dr A Coomarasamy, MRCOG
Reseach Fellow
Birmingham Women's Hospital, UK
I think your concerns are very valid. I think the conventional way of
handling this is to estimate the cost-effectiveness of treatment strategies
involving each diagnostic test, taking into account the sensitivity and
specificity of the tests, the clinical outcomes of all eventualities
including the consequences of actions resulting from false
positives/negatives. This is best done by building a decision tree model to
make each eventuality explicit. Karl Claxton at York ([log in to unmask])
teaches this very nicely in a practical exercise on the MSc course in
health economics, in the context of the test/treat decisions in coronary
artery disease. Specifiying the decision tree in order of the information
seen by the doctor is important - especially converting sensitivity and
specificity values to positive and negative predictive values using Bayes
rule - it's the latter which are needed because what the doctor actually
sees first is the result of the test, not whether the patient actually has
the condition being tested for.
You might look for recent articles by Karl and luminaries in clinical
decision analysis such as Milt Weinstein at Harvard. Also refer to
Weinstein and Fineberg's classic text Clinical Decision Analysis
(Philadelphia, 1980, Saunders), now out of print but I believe Milt has a
second edition in the works. And there was a series on specifiying decision
trees in Medical Decision Making a few years ago. Sorry I don't have
references to hand.
I'd be happy to send you a spreadsheet decision tree model from my student
days at York re the above exercise, if you aren't able to find better.
Regards
Jeremy
Jeremy Chancellor
Managing Director, European Operations
Innovus Research (UK) Ltd
St. Mary's Court
The Broadway
Amersham
Buckinghamshire
HP7 0UT
United Kingdom
You are correct. Ending the evaluation at 'cases detected/diagnosed'
without considering sensitivity and specificity for the economic
consequences of
treatment decision is incomplete. There is an established methodology to
incorporate sensitivity and specificity when the outcomes are modeled in a
decision tree. The method is the treatment-threshold approach. The
following references should help you:
> Detsky AS, Mendelson RA, Baker JP, Jeejeebhoy KN. The choice to treat all,
some
> or no patients undergoing gastrointestinal surgery with nutritional
support: a
> decision analysis approach. Journal of Enteral and Parenteral Nutrition.
> 1984;8:245-253.
>
> Pauker SG, Kassirer JP. The threshold approach to clinical decision
making. New
> England Journal of Medicine. 1980;302:1109-1117.
>
> Textbook: Sox HC, Blatt MA, Higgins MC, Marton KI. Medical Decision
Making.
> Butterworth-Heinemann. Boston 1988. Especially Chapter 9.
>
> I realize these references are quite old. If you come across some more
recent
> papers on the topic I would be most grateful if you would send me the
> references. In your lit searching, you may want to try focusing
specifically on
> economic evalautions that used decision analysis.
> Good luck,
> Wendy Ungar
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> Wendy J. Ungar, M.Sc., Ph.D.
> Scientist, Population Health Sciences
> Assistant Professor, Department of Health Policy, Management and
Evaluation
> Faculty of Medicine, University of Toronto
>
> Division of Population Health Sciences
> Hospital for Sick Children
> 555 University Avenue
> Toronto, Ontario
there are a number of possibilities.
for example, the costs associated with all outcomes could be included in the
cost profile
we have nearly finished a study looking at the use of MRI for staging in
rectal cancer and the same principle applies - ths cost of MRI is much
higher but when you consider its effectiveness and the cosnequences of
incaccurate staging a different set of results emerges
If I can be of further help do not hesitate to get in touch
Phillips, Ceri
I have had exactly the same concerns and am looking for literature
You might like to canb one or two economic evaluation databases, such as DH
HEED at University of York Centre of Reviews and Dissemination. Or try
sending your message to the mailbase healthecon-all
It is important to capture both the long-term benefit of ruling in and
ruling out - the balance. Best to think of economic outcomes for each of the
four cells in the 2x2 table. Especially consider impact of false negatives -
delay of diagnosis in eg. cancer can be serious.
In some cases (?Downs syndrome in pregnancy) the value of information in
itself even though thgis may not lead to further intervention.
NB economic consequences include costs and health outcomes, of course
Need to consider the time horizon over which outcomes should be measured. I
would be grateful if you could send me a summary of the replies you receive.
Mike Chambers
MEDTAP International
I did some work for the NHS Breast Screening Programme a few years ago. A
working group was set up in the '80s to think through all the aspects of
setting up the programme. I believe that an economic analysis was included
in their report. This is rather dated, and I'm not sure of the details of
their analysis. However, this is an area where it would have been worth
considering false positives/negatives, etc. If you think this is worth
following up, it was published in the Forrest Report (Breast Cancer
Screening - Report to the Health Ministers of England, Wales, Scotland &
Northern Ireland by a Working Group Chaired by Professor Sir Patrick
Forrest. HMSO: Department of Health and Social Security, 1986). You may be
able to obtain a copy from the university library. If not, the Breast &
Cervical Quality Assurance Reference Centre is based in Birmingham. When I
was there it was in the New Public Health Building, the University of
Birmingham. I hope this is helpful, sorry if it's going over material
you've already found in your search. All the best for your study.
Isabel Taylor
PhD student
ScHARR
University of Sheffield.
There is no guidance from the literature on how to express the economic
efficiency of a diagnostic test. I recently tackled this issue from a
practical perspective, trying to evaluate the most cost-effective method of
identifying patients for expensive drug therapy. It was an admittedly
extreme case where the high cost of the drug, and its survival gain for
true-positive cases, compelled us to evaluate cost per life year gained.
Neither the "cost per true positive case found", nor the "cost per false
positive case avoided" provides a meaningful ratio. Rather, linking the
true positive cases to patient (expected) outcomes and false positives to
inappropriate costs (i.e., drug utilization) illuminates the economics of
alternative diagnostic methods. It also allows decision makers to compare
diagnostic technologies with other technologies.
If health economics is going to be applied to diagnostics testing, it must
be linked with patient outcomes.
Regards,
Lane Ilersich, MSc BScPhm
Health Outcomes Management
Hoffmann-La Roche Limited (Canada)
> It seems to me that a cost per case diagnosed is reasonable as a surrogate
> (but not ideal) as long as ALL cost are included, including those of
> treating FP's or the additional costs of delaying treatment to FNs, Very
> much as a CEA of a drug would include the costs of treatment failures.
> The most sound way of comparing the tests would of course be in terms of
> their health outcomes, as this is why we are performing the tests in the
> first place, and otherwise we won't capture the full effects of for
instance
> dealying treatment due to an FN.
> Literature on screening programmes often takes this tack, for example
> survival in the case of breast cancer screening.
>
> Mike
>
|