Skip to main content

Optimal Strategy for Multiple Diagnostic Tests

  • Chapter
  • First Online:
Medical Decision Making

Abstract

This chapter deals with the optimal test strategy if multiple diagnostic tests are available and can be employed simultaneously or sequentially. This complicates the task for the decision maker. He has to decide on the positivity criterion for the composite test, which can be conjunctive or disjunctive, and on the order of the tests if he uses them sequentially. If testing is potentially harmful, it is easy to understand that sequential testing always dominates parallel testing.

This is a preview of subscription content, log in via an institution to check access.

Access this chapter

Chapter
USD 29.95
Price excludes VAT (USA)
  • Available as PDF
  • Read on any device
  • Instant download
  • Own it forever
eBook
USD 49.99
Price excludes VAT (USA)
  • Available as EPUB and PDF
  • Read on any device
  • Instant download
  • Own it forever
Softcover Book
USD 64.99
Price excludes VAT (USA)
  • Compact, lightweight edition
  • Dispatched in 3 to 5 business days
  • Free shipping worldwide - see info

Tax calculation will be finalised at checkout

Purchases are for personal use only

Institutional subscriptions

Notes

  1. 1.

    Raiffa, H.—Decision Analysis: Introductory lectures on choices under uncertainty—Addison-Wesley Publishing Company, Reading, Massachusetts—1968, p. 271.

  2. 2.

    A test is (Pareto) efficient if there is no other test that is better than the other in terms of either sensitivity or specificity and that is at the same time not worse than the other.

  3. 3.

    \( L{R}_c^{+}- L{R}_d^{+}>0 \) if \( S{e}_1\cdot S{e}_2\left(1- S{p}_1\cdot S{p}_2\right)-\left( S{e}_1+ S{e}_2- S{e}_1\cdot S{e}_2\right)\left(1- S{p}_1\right)\left(1- S{p}_2\right)>0. \) Transformation leads to \( S{e}_1\cdot S{e}_2\left(1- S{p}_1+1- S{p}_2\right)-\left( S{e}_1+ S{e}_2\right)\left(1- S{p}_1\right)\left(1- S{p}_2\right)>0. \) The next step results in \( S{e}_1\left(1- S{p}_1\right)\left( S{e}_2+ S{p}_2-1\right)+ S{e}_2\left(1- S{p}_2\right)\left( S{e}_1+ S{p}_1-1\right)>0, \) which holds true for tests with discriminatory power.

  4. 4.

    Divide the numerator and denominator of \( L{R}_d^{+} \) by the product \( S{p}_1\cdot S{p}_2 \) to obtain \( L{R}_d^{+}=\left(1/ S{p}_1\cdot S{p}_2- L{R}_1^{-}\cdot L{R}_2^{-}\right)/\left(1/ S{p}_1\cdot S{p}_2-1\right). \) As the negative likelihood ratios are smaller than one, the numerator exceeds the denominator, and hence \( L{R}_d^{+}>1. \)

References

  • Ament, A., & Hasman, A. (1993). Optimal test strategy in the case of two tests and one disease. International Journal of Bio-Medical Computing, 33(3–4), 179–197.

    Article  Google Scholar 

  • Drummond, M. F., O’Brien, B., Stoddart, G. L., & Torrance, G. W. (1987). Methods for the economic evaluation of health care programmes. Oxford: Oxford University Press.

    Google Scholar 

  • Hershey, J. C., Cebul, R. D., & Williams, S. V. (1986). Clinical guidelines for using two dichotomous tests. Medical Decision Making, 6(2), 68–78.

    Article  Google Scholar 

  • Hull, R., Hirsh, J., Sackett, D. L., & Stoddart, G. L. (1981). Cost-effectiveness of clinical diagnosis, venography and non-invasive testing in patients with symptomatic deep-vein thrombosis. New England Journal of Medicine, 304, 1561–1567.

    Article  Google Scholar 

  • Krahn, M. D., Mahoney, J. E., Eckman, M. H., Trachtenberg, J., Pauker, S. G., & Detsky, A. S. (1994). Screening for prostate cancer: A decision analytic view. JAMA, 272(10), 773–780.

    Article  Google Scholar 

  • Perone, N., Bounameaux, H., & Perrier, A. (2001). Comparison of four strategies for diagnosing deep vein thrombosis: A cost-effectiveness analysis. The American Journal of Medicine, 110(1), 33–40.

    Article  Google Scholar 

  • Weinstein, M. C., Fineberg, H. V., Elstein, A. S., Frazier, H. S., Neuhauser, D., Neutra, R. R., et al. (1980). Clinical decision analysis. Philadelphia: W.B. Saunders.

    Google Scholar 

Download references

Author information

Authors and Affiliations

Authors

Exercises

Exercises

  1. 1.

    Assume there is a specific treatment for an illness from which a sick patient will benefit with 0.4 QALYs. But the patient might not actually suffer from the diagnosed illness and therefore experience a loss of 0.1 QALYs from unnecessary treatment. A diagnostic test available to the physician has Se = 0.6 and Sp = 1. Assume that the harm is equivalent to an additional QALY loss equal to l t = 0.01 for both the sick and the healthy. Equation (5.8), then, changes to

    $$ V I(p)=\left\{\begin{array}{l} p\cdot Se\cdot g-\left(1- p\right)\left(1- Sp\right) l-{l}^t\kern0.75em \mathrm{for}\kern1.5em 0\le p<{p}^{Rx},\\ {}- p\cdot g\left(1- Se\right)+\left(1- p\right) Sp\cdot l-{l}^t\kern0.75em \mathrm{for}\kern1.5em {p}^{Rx}\le p\le 1.\end{array}\right. $$
    1. (a)

      Calculate the treatment threshold, the test and test-treatment thresholds, as well as the value of information for the situation in which a non-harmful test is available.

    2. (b)

      Display the situation in a graph.

    3. (c)

      How is the treatment decision affected if a test is harmful?

    A second test now becomes available. It has the same specificity but a higher sensitivity than the first test (Sp = 1 and Se = 0.8). The QALY loss for the second test is 0.02, which is higher than that of the first test.

    1. (d)

      What is the minimum a priori probability of an illness at which the physician will prefer the second test?

    2. (e)

      Calculate the overall sensitivity and specificity if the physician tests sequentially, starting with the first test and proceeding with the second test for patients who received a positive first result.

    3. (f)

      How will the physician’s choice change if the QALY loss of the second test increases to 0.05?

  2. 2.

    Figure 7.3 reveals that one of the two individual tests is always inferior. Prove the following general proposition: If \( V{I}_i(p)> V{I}_c(p) \) or \( V{I}_i(p)> \) VI d (p), then \( V{I}_j(p)< V{I}_i(p), \) \( i\ne j. \)

  3. 3.

    Demonstrate the following two inequalities: \( {p}_d^{tRx}>{p}_c^{tRx}>{p}^{Rx} \). You may follow the approach used in footnotes 2 and 4.

  4. 4.

    Assume two tests with the following properties: \( S{e}_1=0.6,\ S{p}_1=0.9 \) and \( S{e}_2=0.7,\ S{p}_2=0.8. \) Furthermore, the utility gain for sick patients will be 4 QALYs and the utility loss to healthy patients 1 QALY.

    1. (a)

      Calculate the overall sensitivity and specificity of the composite test (for both the conjunctive and the disjunctive positivity criterion).

    2. (b)

      Calculate the positive and negative likelihood ratios of the two single tests, as well as the likelihood ratios of the composite test with conjunctive and disjunctive positivity criteria.

    3. (c)

      Draw Fig. 7.3 for the given test properties.

    4. (d)

      Which test strategy will be preferred at the a priori probabilities of 0.1, 0.5 and 0.9?

    5. (e)

      Draw level curves of informational value for these a priori probabilities.

    6. (f)

      In which prevalence range is which alternative dominant?

    7. (g)

      Draw Fig. 7.6 for the given test properties.

  5. 5.

    Hull et al. (1981) analyzed diagnostic strategies for deep vein thrombosis (DVT). The table below shows the outcomes and costs generated by two alternative strategies: impedance plethysmography (IPG) alone versus IPG plus outpatient venography if IPG is negative.

    Program

    Test result

    Presence of DVT

    Costs (US$)

    yes

    no

    IPG alone

    \( + \)

    142

    5

    321,488

    \( - \)

    59

    272

     

    IPG plus venography

    \( + \)

    201

    5

    603,552

    if IPG is negative

    \( - \)

    0

    272

     

    Perone et al. (2001) report the following (3-month) quality-adjusted expected survival probabilities for a 50-year-old patient with a life expectancy of 29 years: 99.518 % (with DVT and treatment), 97.5 % (with DVT and no treatment), 99.768 % (without DVT and treatment) and 100 % (without DVT and no treatment).

    1. (a)

      Calculate the a priori probability of DVT, the sensitivity and specificity of the individual and the combined test.

    2. (b)

      Calculate the QALY values for each scenario, as well as the utility gain and loss from treatment and the treatment threshold. For the given a priori probability, which is the best strategy without a test?

    3. (c)

      Calculate the test and test-treatment thresholds for each individual test as well as the composite test. For the given a priori probability, which is the best strategy when tests are available?

    4. (d)

      Assume that treatment of a patient costs 2755 US$ and that the shadow value of the budget constraint is 50,000 US$ per additional QALY. Which is the best strategy?

Rights and permissions

Reprints and permissions

Copyright information

© 2017 Springer-Verlag GmbH Germany

About this chapter

Cite this chapter

Felder, S., Mayrhofer, T. (2017). Optimal Strategy for Multiple Diagnostic Tests. In: Medical Decision Making. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-662-53432-8_7

Download citation

  • DOI: https://doi.org/10.1007/978-3-662-53432-8_7

  • Published:

  • Publisher Name: Springer, Berlin, Heidelberg

  • Print ISBN: 978-3-662-53431-1

  • Online ISBN: 978-3-662-53432-8

  • eBook Packages: Economics and FinanceEconomics and Finance (R0)

Publish with us

Policies and ethics