MLS 3302 BIOSTATISTICS RESEARCH METHODS LABORATORY PRACTICES Unit
MLS 3302 BIOSTATISTICS, RESEARCH METHODS, & LABORATORY PRACTICES Unit 3 – Laboratory Statistical Applications Section 2 – Predictive Values Matthew Nicholaou, Dr. PH, MT(ASCP)
Unit 3 – Overview Section 1 � Section 2 � Reference Ranges Section 5 � Accuracy and Precision Section 4 � Predictive Values Section 3 � Pre-Use Considerations Method Comparison Section 6 � Minimum Detection Limit
Unit 3. 2 – Predictive Value Theory One of the key functions of the clinical laboratory is to provide information to aid the diagnosis of disease Some diagnoses are Qualitative � Strep throat, Mono Some diagnoses are Quantitative � Dyslipidemia
Unit 3. 2 – Predictive Value Theory Quantitative tests � Have defined Reference Ranges that classify people as healthy or unhealthy (diseased) � How to define reference ranges will be discussed later, but generally a healthy range contains ~95% of values in people that do not have a disease or pathology
Unit 3. 2 – General Terms It is common to use predictive values to compare a new qualitative test to an existing Gold Standard True Positive (TP) – subject has a disease and a positive result on the new test True Negative (TN) – subject does not have a disease and a negative result on the new test
Unit 3. 2 – General Terms False Positive (FP) – subject does not have the disease but tests positive on the new test False Negative (FN) – subject does have the disease but test negative on the new test Note: we have to know the true disease status of the patient, usually done by performing the gold standard test
Unit 3. 2 – Sensitivity and Specificity Analytic � Sensitivity – lowest level of a substance that can be detected (minimum detection limit) � Specificity – how well an assay detects a specific substance Diagnostic � Sensitivity – the ability of a test to identify positive results (influenced by FN) � Specificity – the ability of a test to identify negative results (influenced by FP)
Unit 3. 2 – Sensitivity and Specificity The best new test would ideally have a high diagnostic sensitivity and specificity… but often there is an inverse relationship between the two. That is, as you increase the sensitivity of a test the specificity will decrease and visa versa.
Unit 3. 2 – Positive and Negative Predictive Values Positive Predictive Value (PPV) – the proportion of positive results that are true positives � Influenced by FP Negative Predictive Value (NPV) – the proportion of negative results that are true negatives � Influenced by FN
Unit 3. 2 – 2 x 2 Table New Test Gold Standard Positive Negative Positive TP FP (Type I Error) PPV TP / (TP + FP) Negative FN (Type II Error) TN NPV TN / (FN + TN) Sensitivity TP / (TP + FN) Specificity TN / (FP + TN)
Unit 3. 2 - Efficiency – The test’s ability to correctly identify both true positive and true negative individuals in the population % Efficiency = [(TP + TN) / Total Sample] x 100
Unit 3. 2 - Example Scenario A company has developed a new rapid screen test to detect Group A Strep They asked you to evaluate this test in your hospital over the next peak strep season Test results were compared to the ‘gold standard’ culture method
Unit 3. 2 - Example Results A strep screen and culture were performed on 1, 000 patient samples over three months 540 patients tested positive on the screen � of which, 500 were positive on culture 460 patients tested negative on the screen � of which, 400 were negative on culture
Unit 3. 2 – Example Strep Screen Culture Positive Negative Positive 500 40 Negative 60 400
Unit 3. 2 – 2 x 2 Table New Test Gold Standard Positive Negative Positive TP FP (Type I Error) PPV TP / (TP + FP) Negative FN (Type II Error) TN NPV TN / (FN + TN) Sensitivity TP / (TP + FN) Specificity TN / (FP + TN)
Unit 3. 2 – Example Strep Screen Culture Sensitivity Specificity PPV NPV Efficiency Positive Negative Positive 500 40 Negative 60 400 = TP / (TP + FN) = 500 / (500 + 60) = 0. 892 x 100 = 89. 2% = TN / (TN + FP) = 400 / (400 +40) = 0. 909 x 100 = 90. 9% = TP / (TP + FP) = 500 / (500 + 40) = 0. 926 x 100 = 92. 6% = TN / (TN + FN) = 400 / (400 + 60) = 0. 869 x 100 = 86. 9% = (TP + TN) / Total = (500 + 400)/1000 = 0. 90 x 100 = 90%
Unit 3. 2 – Example Sensitivity = 89. 2% Specificity = 90. 9% PPV = 92. 6% NPV = 86. 9% Efficiency 90% This screen performs really well. It is better for a screen of a treatable and less severe disease to have FPs because usually they are confirmed with the gold standard. With more ‘severe’ diseases, e. g. HIV, it is still better to falsely identify people rather than miss positive individuals but there is a need to maximize sensitivity to minimize any harm to FPs
Unit 3. 2 – Considerations Generally predictive values are used to assess new screening tests The validity is highly dependent on sample size, almost always should be done on a very large sample… ESPECIALLY if the disease prevalence is low
Unit 3. 2 – Predictive Values
- Slides: 19