Professional Documents
Culture Documents
By Carol Lee
Copyright 2002-2010, Data Innovations, LLC. All rights reserved world-wide. This document may not be reproduced without express written permission.
Principle
Qualitative methods and semi-quantitative methods lack sufficient analytical sensitivity to demonstrate accuracy across a wide range of observed results. Therefore: The results are categorized in states of diagnostic description. i.e.: postive / negative, equivocal, gray zone, Qualitative 2 states (i.e. positive and negative) Semi-quantitative a limited number of states (up to 6 are allowed in EP Evaluator).
2
QMC
Evaluate the degree of concordance between two qualitative or
semi-quantitative methods.
colorful bubble chart size of bubbles proportional to number of points concordance or truth tables
Experimental Design
Qualitative:
At least 20 specimens. Equal numbers of positives and negatives
Semi quantitative
Same number of states for both methods User defined results or cutoffs and labels 0 to 10, few, Try for similar N in each state.
Calculations
If One method is specified as a gold standard (i.e. absolutely correct). calculates specificity and sensitivity If Neither method is a gold standard. Only calculates relationship of the two methods, namely the degree of agreement and whether one method is more or less sensitive relative to the other.
Qualitative
Default is 2 states P = positive N = Negative Allowed results show in the drop down box in the data entry area
Gold standard
Report Excerpt
Prepared for: chemistry Dept -- Holy Name hospital By: Clinical Laboratory -- Community Hospital
Statistical Analysis
(Comparison of two Laboratory Methods)
McNemar Test for Symmetry: Test < Reference 14 (2.6%) Test > Reference 15 (2.8%) Symmetry test PASSES p = 0.853 (ChiSq=0.034, 1 df)
A value of p<0.05 suggests that one method is consistently "larger".
Cohen's Kappa
Kappa is the proportion of agreement above what's expected by chance. Rule of thumb is Kappa>75% indicates "high" agreement. We would like to see VERY high (close to 100%) agreement.
Statistical Summary
Negative Reference Negative Test
222
Positive Reference
14
Legend:
Total
236 1 2 Negative (N) Positive (P) Negative (N) Positive (P)
Concordance table
result accurately shows the absence of the disease. Agreement: The percent of total cases in which the two methods give the same result. Related statistics for qualitative tests: Positive Agreement is the percent of cases that match when the reference method is positive: TP/(TP+FP). Negative Agreement is the percent of cases that match when the reference method is negative: TN/(TN+FN). Cohen's Kappa: Similar to Agreement, but adjusted for the probability that the two methods agree by chance. Kappa ranges from -100% to 100%. A value of zero indicates random agreement. A value of 100% indicates perfect agreement. It is desirable for Kappa to be well above 75%. McNemar Test for Symmetry: A test for bias -- whether one method is consistently larger than the other. If the number of cases where X>Y is equal (within random error) to the number of cases X<Y, the method is unbiased, and the symmetry test passes. If most of the differences between X and Y occur when X>Y (or when X<Y), the symmetry test fails.
Statistical tests
parison
non-quantitative stick method for he results for the ps corresponding to n be used to h quantitative tests used in nt.
ely correct.
s positive by both
Preliminary Report
The word PRELIMINARY printed diagonally across the report indicates that the data is incomplete, and the report is not acceptable as a final report. Some or all of the statistics may be missing. This report is preliminary if there are less than 20 unexcluded results.
s negative by both
Semi quantitative
Prepared for: chemistry Dept -- Holy Name hospital By: Clinical Laboratory -- Community Hospital
Statistical Analysis
6 5 4 3 2 1
(Comparison of two Laboratory Methods)
Agreement
McNemar Test for Symmetry: Test < Reference 23 (25.8%) Test > Reference 2 (2.2%) Symmetry test FAILS p < 0.001 (ChiSq=17.640, 1 df)
A value of p<0.05 suggests that one method is consistently "larger".
Cohen's Kappa
1 2 3 4 5 neg <-- Reference -->pos 6
Kappa is the proportion of agreement above what's expected by chance. Rule of thumb is Kappa>75% indicates "high" agreement. We would like to see VERY high (close to 100%) agreement.
Test
Reference 1 2 3
Legend:
4 5 6 Total Reference Test
Allow 1 step
Prepared for: chemistry Dept -- Holy Name hospital By: Clinical Laboratory -- Community Hospital
Statistical Analysis
6 5 4 3 2 1
(Comparison of two Laboratory Methods)
McNemar Test for Symmetry: Test < Reference 23 (25.8%) Test > Reference 2 (2.2%) Symmetry test FAILS p < 0.001 (ChiSq=17.640, 1 df)
A value of p<0.05 suggests that one method is consistently "larger".
Cohen's Kappa
1 2 3 4 5 neg <-- Reference -->pos 6
Kappa is the proportion of agreement above what's expected by chance. Rule of thumb is Kappa>75% indicates "high" agreement. We would like to see VERY high (close to 100%) agreement.
Test 1
Reference 1 2 3
10 5 --
Legend:
4
--
5
--
6
--
Total
15 1
Reference
Very Negative
Test
Very Negative
Jan 6, 2010
17
LvlN Analyt Custom Result ame LvlVal LvlNa LvlVal LvlNa LvlVal LvlNa LvlVal LvlNa LvlVal LvlNa LvlVal e ? sType 1 ue1 me2 ue2 me3 ue3 me4 ue4 me5 ue5 me6 ue6 BACT y fuzzy cells y
Jan 6, 2010
Nor 1m Nor 1m
Abn Abn
1200 1
18
Jan 6, 2010
19
20