Wallach's Interpretation of Diagnostic Tests: Pathways to Arriving at a Clinical Diagnosis Read Online Free PDF

on the ROC curve closest to the far upper left corner (x = 0, y = 1).
    However, finding the right balance between optimal sensitivity and specificity may not involve simultaneously minimizing false positives and false negatives in all situations. For example, when screening for a deadly disease that is curable, it may be desirable to accept more false positives (lower specificity) in return for fewer false negatives (higher sensitivity). ROC curves allow a more thorough evaluation of a test and potential cutoff values but are not the ultimate arbiters of how to set sensitivity and specificity.
    POSTANALYTIC ERRORS
    Approximately 70–80% of the patient chart or medical record is composed of laboratory test results. Postanalytic errors are dependent on the design and development of those processes and procedures that will ensure correct and timely notification of these test results to the patient’s medical record with right reference range and appropriate interpretation of the test result. Manual and telephone reporting should be discouraged as this reporting is subject to transcription errors at the receiver end. The introduction of a hospitalized computer order entry system has eliminated some errors, but it has not eliminated the risk of mismatching the patients.
    REFERENCE INTERVALS
    The term “reference values” has essentially replaced the obsolete term “normal values.” Laboratory tests are commonly compared to a reference interval before health care providers make physiologic assessments, medical diagnosis, or management decisions. These comparisons may be cross-sectional or longitudinal. A cross-sectional comparison is comparison of an analyte result for a single patient with the interval of results for that analyte obtained from a group of apparently healthy individuals. This is referred to as the “population-based” reference interval. Another example of a cross-sectional comparison is when a single patient result is compared with a fixed value or cutoff value. There are two types of population-based reference intervals. The most common type is derived from a reference sample of persons who are in good health (health associated). The other type of reference interval has been termed “decision based” and defines specific medical decision limits that clinicians use to diagnose or manage patients. Longitudinal comparisons are when a patient’s most recent value is compared with previous values for the same analyte. This may help detect a change in health status.
    Comparison of patient results with a population-based reference interval or with the cutoff values is used for diagnostic or screening purposes. The reference value change over a period of time is used for monitoring patients. Both healthy reference limits and disease-associated reference limits are important for the clinical interpretation of the laboratory test results and vary from laboratory to laboratory. These variations may be caused by preanalytic processing procedures, populations of healthy individuals, inherent random biologic variations, analytic platforms, or analytic imprecision that was present when reference intervals were determined.
    Decision limits for optimally classifying patients into “disease” versus “healthy” categories are difficult to define. Most diseases are not homogenous distributions but represent a continuum of mild and severe forms. Various statistical tools and models have been developed to formalize the medical decision process, but most of the models do not include the methodologic differences in laboratory test values. The major utility of healthy reference intervals for clinicians is to provide a rough assessment of the possibility that a test value on a specific patient is difficult for the values normally found in similar healthy subjects. The guidelines for medical decision making use a standard 95% reference interval. By defining the healthy reference interval to include central 95% of