**KEY** Chapter 4: Do not share without written permission!

 

4.1 - 4.6 See text.

 

4.7  kappa = 0.726; excellent agreement.

 

4.8 SEN = 15 / 18 = 0.833                SPEC = 145 / 152 = 0.954

 

4.9 Exercise 4.7 is a reproducibility analysis because it considers agreement between two raters, neither of which is a gold standard. 

Exercise 4.8 is a validity analysis because it gauges the results to what actually is.

 

4.10 kappa = 0.56; good agreement. 

 

4.11  PVP = 11 / 25 = 0.4400

 

4.12 Screening for bladder cancer

 

(A) 2-by-2 table showing the number of TPs, TNs, FNs, and FPs 

D+

D-

Total

T+

400

3980

4380

T-

100

95,520

95,620

Total

500

99,500

100,000

 

(B) How many people will have a positive test result?      4380   Of these, what proportion will be true positives?   400 / 4380 = 0.091 (9.1%) 

 

(C) How many people will have a negative test result?  95,620   Of these, what proportion will be true negatives?  95520 / 95620 = 0.999 (99.9%)

 

(D) If the SENsitivity of the test were increased from 80% to 90% (keeping SPEC at 96%):

Test

D+

D-

Total

D+

450

3980

4430

D-

50

95520

95570

Total

500

99500

100000

PVP = 450 / 4430 = 0.102 (10.2%)

 

(E) If the SPECificity were were increased from 96% to 98% (keeping SEN at 80%), what would the PVP of the test be?

Test

D+

D-

Total

T+

400

1990

2390

T-

100

97510

97610

Total

500

99500

100000

PVP = 400 / 2390 = 0.167 (16.7%) 

 

(F) Based on your answers to (G) and (H), what has greater influence on the PVP of the test, increasing the SEN or increasing the SPEC? Increasing SPEC.  Does either result in a test that has good predictive value? Neither modification resulted in a test with adequate predictive value. If not, what practice can effectively increase the predictive value of a positive test?  Multiple levels of screening may be necessary.

Last update: 05/21/2009