More than 80% of studies fulfilled the standards for reporting indeterminate or uninterpretable test results, but only 67% of articles indicated how indeterminate results were incorporated into measures of test accuracy. Exclusion of indeterminate or uninterpretable results from analysis results in considerable bias unless they occur in a random manner. Studies also commonly failed in 83% of instances to provide information on the reproducibility of diagnostic test results in detail canadian neightbor pharmacy. Tests that had objective results (eg, biochemical tests) provided information on the reproducibility and variability of measured values in only 18% of reports. Tests that required subjective interpretation (eg, imaging studies) provided information on interobserver agreement in only 15% of studies. Absence of this information severely limits the evaluation of test feasibility and applicability in different clinical settings.
Most of the study articles missed an opportunity to provide a comprehensive assessment of the diagnostic accuracies of the evaluated tests. Seventeen percent (17%) of the articles did not provide information in the form of any of the standard measures of discriminative properties listed in Table 3. More than 80% of articles reported values for sensitivity and specificity, and these measures were the only values provided for diagnostic accuracy in 17% of the articles. Reliance on these measures is unfortunate considering the limitations of sensitivity and specificity for describing the diagnostic value of a new test. Sensitivity and specificity are subject to spectrum bias, which is an important consideration in this study because case mix was defined in only 61% of articles. Also, sensitivity and specificity have less relevance to clinical decision-making than other measures of diagnostic accuracy because they do not describe the likelihood that a patient with a given test result has the target disease. Predictive values, likelihood ratios, odds ratios, and ROC analyses provide more information as general measures of the overall accuracy and clinical utility of a diagnostic test. These indexes were employed by only a small proportion of reviewed articles. Also, few studies provided confidence intervals for the measures of the tests’ diagnostic properties; these values are necessary for determining the precision by which investigations established the accuracy of a test.