Presentation is loading. Please wait.

Presentation is loading. Please wait.

Epidemiology Kept Simple Chapter 4 Screening for Disease.

Similar presentations


Presentation on theme: "Epidemiology Kept Simple Chapter 4 Screening for Disease."— Presentation transcript:

1

2 Epidemiology Kept Simple Chapter 4 Screening for Disease

3 §4.1 Introduction (comment) Accurate identification of true health status -- essential for all epi. workAccurate identification of true health status -- essential for all epi. work The term disease refers to any health condition or health determinantThe term disease refers to any health condition or health determinant

4 Sources of Epi Information Interview (direct, surrogate)Interview (direct, surrogate) QuestionnaireQuestionnaire Pre-existing recordsPre-existing records Direct examinationDirect examination

5 Signs, Symptoms, and Tests Sign = observation by trained observerSign = observation by trained observer Symptom = observation by patientSymptom = observation by patient Tests = any objective resultTests = any objective result

6 Reproducibility and Validity Reproducibility = agreement upon repetitionReproducibility = agreement upon repetition Validity = ability to discriminate those with and without diseaseValidity = ability to discriminate those with and without disease A test can be reproducible but not validA test can be reproducible but not valid A test can be valid but not reproducibleA test can be valid but not reproducible

7 §4.2 Reproducibility Statistics Overall agreementOverall agreement Percentage of diagnoses that agreePercentage of diagnoses that agree Some agreement due to chance (e.g., two coins flipped simultaneously will agree 50% of time)Some agreement due to chance (e.g., two coins flipped simultaneously will agree 50% of time) Kappa statistic - quantifies percent agreement above chanceKappa statistic - quantifies percent agreement above chance

8 Kappa Statistic (Formula) Rater 2 Rater 1 +-Total +ab p1p1p1p1 -cd q1q1q1q1 Total p2p2p2p2 q2q2q2q2N Text says to convert counts to proportions, but this is not necessary

9 Kappa Statistic (Illustration) Rater 2 Rater 1 +-Total +20424 -57176 Total2575100

10 Interpretation of Kappa Percent agreement above chancePercent agreement above chance The closer to 1, better agreementThe closer to 1, better agreement Range of Kappa Interpretation >.75 Excellent agreement.40 to.75 Good agreement <.40 Poor agreement

11 §4.3 Validity Compare test results to definitive diagnostic procedure (“gold standard”)Compare test results to definitive diagnostic procedure (“gold standard”) Each case classified asEach case classified as TP = true positiveTP = true positive TN = true negativeTN = true negative FP = false positiveFP = false positive FN = false negativeFN = false negative

12 Organize Data in Table Gold Standard Test Result +-Total +TPFPTP+FP -FNTNFN+TN TotalTP+FNFP+TN n

13 Definitions Notation: Pr(T|D) = probability test result given disease statusNotation: Pr(T|D) = probability test result given disease status Sensitivity = Pr(T+|D+)Sensitivity = Pr(T+|D+) Specificity = Pr(T-|D-)Specificity = Pr(T-|D-) Predictive value positive = Pr(D+|T+)Predictive value positive = Pr(D+|T+) Predictive value negative = Pr(D-|T-)Predictive value negative = Pr(D-|T-) Prevalence = Pr(D+)Prevalence = Pr(D+)

14 Formulas SEN = TP / (TP + FN)SEN = TP / (TP + FN) SPEC = TN / (TN + FP)SPEC = TN / (TN + FP) PVP = TP / (TP + FP)PVP = TP / (TP + FP) PVN = TN / (TN + FN)PVN = TN / (TN + FN)

15 Example (HIV Screening Test) Prevalence = 1000 / 1,000,000 =.001 SEN = 990 / 1000 =.99SEN = 990 / 1000 =.99 SPEC = 989,010 / 999,000 =.99SPEC = 989,010 / 999,000 =.99 PVP = 990 / 10,980 =.09PVP = 990 / 10,980 =.09 PVN = 989,010 / 989,020  1.000PVN = 989,010 / 989,020  1.000 HIV Screening Test +-Total +9909,99010,980 -10989,010989,020 Total1000999,0001,000,000

16 Example (HIV Screening Test) Prevalence = 10,000 / 1,000,000 =.10 SEN = 99000 / 100,000 =.99SEN = 99000 / 100,000 =.99 SPEC = 891,000 / 900,000 =.99SPEC = 891,000 / 900,000 =.99 PVP = 99,000 / 108,000 =.92PVP = 99,000 / 108,000 =.92 PVN = 891,000 / 900,000 =.99PVN = 891,000 / 900,000 =.99 HIV Screening Test +-Total +99,0009,000108,000 -1,000891,000892,000 Total100,000900,0001,000,000

17 Conclusion A test that is 99% sensitive and 99% specific is used in two populationsA test that is 99% sensitive and 99% specific is used in two populations In low prevalence population PVP =.09In low prevalence population PVP =.09 In high prevalence population, PVP =.92In high prevalence population, PVP =.92 Therefore, PVP is a function of prevalenceTherefore, PVP is a function of prevalence

18 Relation Between Predictive and Prevalence Value PVP is a function ofPVP is a function of Test SENTest SEN Test SPECTest SPEC prior probability (prevalence) of diseaseprior probability (prevalence) of disease Bayesian formulas on pp. 88 - 92 (NR)Bayesian formulas on pp. 88 - 92 (NR)

19 Selecting a Cutoff for a Test HIV screening test detects color change (Optical Density Ratio)HIV screening test detects color change (Optical Density Ratio) Can change cutoff for amount of color change to classify as “positive”Can change cutoff for amount of color change to classify as “positive”

20 Use Low Cutoff no false negatives (high SEN)no false negatives (high SEN) many false positive (low SPEC)many false positive (low SPEC)

21 Use High Cutoff no false positive (high SPEC)no false positive (high SPEC) many false negative (low SEN)many false negative (low SEN)

22 Use In-Between Cutoff medium no. of false positive (intermediate SPEC)medium no. of false positive (intermediate SPEC) medium no. of false negative (intermediate SEN)medium no. of false negative (intermediate SEN)


Download ppt "Epidemiology Kept Simple Chapter 4 Screening for Disease."

Similar presentations


Ads by Google