Abstract/Details

Some approaches to quality in the presence of inspection error: With application to optimal laboratory cancer screening policies


1997 1997

Other formats: Order a copy

Abstract (summary)

This research addresses several quality control problems which arise in a variety of manufacturing, healthcare, service, finance, and other industries given the existence of human and automated attribute detection error. Several mathematical and economic models are developed for various types of single and multiple inspection screening policies in order first to examine inherent tradeoffs between type I errors, type II errors, and all associated inspection, false-rejection, and false-acceptance costs and then, ultimately, to help identify the minimum expected total cost policy and the optimal amount of inspection for any particular scenario. While originally motivated by industrial problems, these models also have been adapted to various non-manufacturing concerns, including service processes and laboratory cancer screening policies. In particular, similar methods are developed and used to analyze the policy for screening Pap smears for early indications of cervical cancer currently required by the Congressional Laboratory Improvements Amendments Act of 1988 (CLIA'88), to compare this policy with possible alternatives, and to develop an algorithm that identifies the optimal policy in any given scenario.

Results show that the mandated CLIA policy never is optimal and always increases total costs, that overall sensitivity of CLIA never can be improved beyond a certain mathematical bound, that CLIA's 10% minimum requirement nor any other amount of partial resampling ever is optimal, that multiple readings in some realistic cases can result in very significant benefits, and that the proposed use of automated rescreening technology recently approved by the FDA may not result in improvements over CLIA nor the optimal policy derived here. Sensitivity analyses indicate that the improvement possible by switching to this optimal policy ranges from 90,000 to 165,000 fewer false-negatives and $250 million to \$1 billion savings annually nationwide. Similar results and savings are demonstrated in several industrial and service applications, with Deming's $k\sb1/k\sb2$ minimum cost criterion in some cases now resulting in the maximum expected cost policy and with multiple inspections often being economically optimal. Several directions for further work are suggested, including extending results to mammography and other cancer screening tests, and an extensive bibliography is provided.

Indexing (details)


Subject
Industrial engineering;
Pathology;
Operations research
Classification
0546: Industrial engineering
0571: Pathology
0796: Operations research
Identifier / keyword
Health and environmental sciences; Applied sciences
Title
Some approaches to quality in the presence of inspection error: With application to optimal laboratory cancer screening policies
Author
Benneyan, James Christian
Number of pages
426
Publication year
1997
Degree date
1997
School code
0118
Source
DAI-B 58/09, Dissertation Abstracts International
Place of publication
Ann Arbor
Country of publication
United States
ISBN
9780591597554, 0591597551
Advisor
Seiford, Lawrence M.
University/institution
University of Massachusetts Amherst
University location
United States -- Massachusetts
Degree
Ph.D.
Source type
Dissertations & Theses
Language
English
Document type
Dissertation/Thesis
Dissertation/thesis number
9809306
ProQuest document ID
304376998
Copyright
Database copyright ProQuest LLC; ProQuest does not claim copyright in the individual underlying works.
Document URL
http://search.proquest.com/docview/304376998
Access the complete full text

You can get the full text of this document if it is part of your institution's ProQuest subscription.

Try one of the following:

  • Connect to ProQuest through your library network and search for the document from there.
  • Request the document from your library.
  • Go to the ProQuest login page and enter a ProQuest or My Research username / password.