Agreement & Result Code Guidelines
Agreement Statistics
The module computes the following statistics to characterize how consistently the two methods classify specimens:
Overall Percent Agreement (OPA)
OPA = (Number of specimens with identical results on both methods / Total specimens) × 100
The proportion of specimens for which both methods assign the same categorical result. OPA is a convenient summary measure of concordance. The module reports OPA with a 95% confidence interval calculated using the Wilson score method, which is recommended by CLSI EP12 and performs well for small sample sizes and proportions near 0% or 100%.
Positive Percent Agreement (PPA)
PPA = A / (A + C) × 100, where A = both methods positive, C = reference positive / test negative
The proportion of reference-positive specimens also classified as positive by the test method. PPA is the appropriate term when the comparator is not a widely accepted best available assessment of the target condition (i.e., not a true gold standard). When the reference method is a recognized gold standard, the equivalent statistic is called clinical sensitivity.
Negative Percent Agreement (NPA)
NPA = D / (B + D) × 100, where D = both methods negative, B = reference negative / test positive
The proportion of reference-negative specimens also classified as negative by the test method. NPA is the appropriate term when the comparator is not a widely accepted gold standard. When the reference method is a recognized gold standard, the equivalent statistic is called clinical specificity. Together, PPA and NPA are the primary agreement metrics recommended by CLSI EP12.
McNemar Test for Symmetry (Supplemental)
χ² = (|b − c| − 1)² / (b + c), where b = Test Positive / Reference Negative, c = Test Negative / Reference Positive
A chi-square test that evaluates whether the disagreements between the two methods are symmetric — that is, whether the test method is equally likely to over-call and under-call relative to the reference method. A p-value > 0.05 (PASS) indicates no statistically significant asymmetry in disagreements. A p-value ≤ 0.05 suggests that one method systematically classifies more specimens as positive than the other. Note: McNemar's test does not appear anywhere in CLSI EP12. It is an established supplemental statistical tool used in the broader diagnostic accuracy literature and by some software platforms.
Cohen's Kappa (κ) (Supplemental)
κ = (Observed Agreement − Expected Agreement) / (1 − Expected Agreement)
A chance-corrected measure of agreement that accounts for the possibility that the two methods might agree by random chance alone, given the prevalence of positives and negatives in the study sample. Kappa is a coefficient from −1 to 1, not a percentage: 1 is perfect agreement, 0 is no agreement beyond chance, and a negative value is less agreement than chance alone would produce. The module reports Kappa to four decimal places (for example, 0.8750) with a 95% confidence interval. Values ≥0.80 are generally interpreted as strong agreement; values 0.60–0.79 as moderate agreement; values below 0.60 warrant investigation. Note: Cohen's Kappa does not appear anywhere in CLSI EP12. The kappa statistic (κ) is briefly mentioned in EP12 Appendix G only as historical context for why average positive agreement (APA) and average negative agreement (ANA) were proposed as alternatives. Kappa is widely used in the broader diagnostic accuracy literature and by some software platforms as a supplemental metric.
The module also reports directional discordance counts:
Test > Reference
Specimens where the test method assigned a higher-level result than the reference method (e.g., test called Positive when reference called Negative). Indicates possible over-calling by the test method.
Test < Reference
Specimens where the test method assigned a lower-level result than the reference method (e.g., test called Negative when reference called Positive). Indicates possible under-calling by the test method.
Custom Result Codes
By default, the module uses simple Positive/Negative result coding. Custom result codes allow you to define between 2 and 6 result levels to accommodate lab-specific terminology or multi-level reporting schemes. Two coding approaches are supported:
Categorical Results
Used when the test reports non-numeric category codes (e.g., +1, +2, Neg, Reactive). You define each result code and assign it a report label — the human-readable name displayed in reports (e.g., 'Non-Reactive', 'Weak Positive', 'Reactive'). The result codes entered must match the actual values in your data.
Numeric Cutoff
Used when the test reports a numeric value (e.g., an index or absorbance ratio) that is interpreted against one or more thresholds to produce a categorical result. You specify the primary cutoff value, whether a number above or below that cutoff represents a positive result, and a report label for each defined level. The first and last boundary values are determined automatically from the cutoff and the number of levels.
The same coding configuration is applied independently to both the reference method and the test method, allowing each to use different cutoff values or categorical schemes if the methods differ in their result reporting.