UncertainInterval: Set of functions for the determination of an Uncertain...

In UncertainInterval: Uncertain Interval Methods for Three-Way Cut-Point Determination in Test Results

Description

A collection of functions to determine a range of test scores that are inconclusive and do not allow a diagnosis (other than Uncertain) and to access its qualities.

Details

Uncertain test scores are scores that have about the same density in the two distributions of patients with and without the targeted condition. This range is typically found around the optimal cut-point, that is the point of intersection or Youden index (Schisterman et al., 2005).

From version 0.7 onward, tests with lower values that indicate the presence of the targeted condition can be analyzed without having to take negative values for the test. Version 0.7 also corrects four bugs (see Github UncertainInterval Issues). Lastly, version 0.7 introduces four ShinyApps that allow for hands-on simulation of the methods ui.binormal, TG.ROC, grey-zone and ROC. These examples are documented within the ShinyApp's and show how they work with a large variety of different tests of varying qualities.

Criteria

Most functions in this package use a specified low value for the sensitivity and specificity of the test scores within the uncertain interval to find this uncertain interval (default UI.Se = UI.Sp = .55). The most recent added function RPV for ordinal test scores uses the odds of the target condition of near 1 to identify the uncertain interval (default < 2). This library also contains two alternative definitions. 1. Coste et al. (2003) defined a grey zone in between positive and negative conclusions (see greyzone), minimum desired values for respectively the positive and negative post-test probability, with defaults .95 and .05. 2. Greiner (1995) defined a middle inconclusive zone of intermediate values (see TG.ROC), with desired minimum values for dichotomous Se and Sp, with default values of .9. See Index for all available functions and plot possibilities.

Glossary

In general, the prefix MCI is used when a statistic is calculated for the test scores that are used for a positive or negative classification. The prefix UI is used when the statistic is applied to the test scores in the uncertain interval.

• Se and SpSe and Sp are statistics that are developed for a single dichotomous cut-point.

• MCI.Se and MCI.Sp Sensitivity and specificity calculated for the More Certain Intervals (MCIs) outside the Uncertain Interval (UI), that is, omitting the test scores in the UI. The meaning of Se and Sp changes from sensitivity and specificity of the test (or all test scores) to sensitivy and specificity of the test scores used for classification.

• UI.Se and UI.SpSensitivity and specificity for the test scores inside the uncertain interval. Please note that the uncertain interval always falls around the point of intersection (optimal threshold or Youden threshold) and that for the calculation of UI.Se and UI.Sp the point of intersection is used as threshold within the uncertain interval.

• NPV and PPVPredictive values for respectively the negative and the positive class. Can be used with both dichotomous and trichotomous sections of the test scores. The prefix MCI is sometimes used, but is superfluous.

• PV.classPredictive value for class when the meaning of class is selfexplanatory.

• NPV.class and PPV.class Negative and Positive Predictive value when the scores in class are used for a negative, respectively positive classification. When predictive values are calculated for the same class, NPV.class = 1 - PPV.class.

• NPV.ui and PPV.ui Negative and Positive Predictive value when all test scores in the uncertain interval would be used for a negative, respectively positive classification. These values can be expected to be close to .5. When predictive values are calculated for the same class, NPV.ui = 1 - PPV.ui.

• UI.NPV and UI.PPVNegative and Positive Predictive value when test scores in the uncertain interval respectively above and below the point of intersection would be used for a negative, respectively positive classification. These values can be expected to be close to .5, but slightly higher than NPV.ui and PPV.ui

• SNPV, SPPV, SPV.class, SNPV.class, SPPV.class, SNPV.ui, SPPV.ui, UI.SNPV and UI.SPPVThe standardized versions of the predictive values mentioned above.

References

Landsheer, J. A. (2016). Interval of Uncertainty: An Alternative Approach for the Determination of Decision Thresholds, with an Illustrative Application for the Prediction of Prostate Cancer. PloS One, 11(11), e0166007.

Landsheer, J. A. (2018). The Clinical Relevance of Methods for Handling Inconclusive Medical Test Results: Quantification of Uncertainty in Medical Decision-Making and Screening. Diagnostics, 8(2), 32. https://doi.org/10.3390/diagnostics8020032

Schisterman, E. F., Perkins, N. J., Liu, A., & Bondell, H. (2005). Optimal cut-point and its corresponding Youden Index to discriminate individuals using pooled blood samples. Epidemiology, 73-81.

Greiner, M. (1995). Two-graph receiver operating characteristic (TG-ROC): A Microsoft-EXCEL template for the selection of cut-off values in diagnostic tests. Journal of Immunological Methods, 185(1), 145-146.

Coste, J., & Pouchot, J. (2003). A grey zone for quantitative diagnostic and screening tests. International Journal of Epidemiology, 32(2), 304-313.

See Also

ui.nonpar, plotMD, get.intersection, quality.threshold, quality.threshold.uncertain

UncertainInterval documentation built on March 3, 2021, 1:10 a.m.