Lecture 8 (Measurement validity, reliability, screening) Flashcards Preview

Quant Research > Lecture 8 (Measurement validity, reliability, screening) > Flashcards

Flashcards in Lecture 8 (Measurement validity, reliability, screening) Deck (17):
1

Measures of disease (3)

-Diagnostic test (symptoms)
-Screening test (absence of symptoms)
-Measurement in epi study

2

Validity of measure of disease/exposure

-Degree to which it distinguishes btwn those who are affected/those who aren't
-Exposure: distinguishing among diff levels

3

Sensitivity

% of ppl w/disease who test +
% of exposed who test says exposed

4

Specificity

% of ppl w/o disease who test -
% non-exposed who test says non-exposed

5

Positive predictive value

% of ppl w/a + test who have disease
% who test + for exposure who actually are exposed

6

Negative predictive value

% of ppl w/a - test who don't have disease
% who test - for exposure who are actually non-exposed

7

Screening tests (why, which diseases)

-Absence of symptoms
-Secondary prevention
-Not diagnostic

When:
-Diseases w/sig mortality
-Treatment has advantage
-Test = simple, fast, cheap, safe

8

Low value cut point (screening test) -->

high sensitivity, low specificity

9

High value cut point (screening test) -->

high specificity, low sensitivity

10

Receiver operating characteristic curve

-plot of sensitivity vs 1 - specificity
-Shows optimum cut point (shoulder of curve), mid point

11

Cost of low sensitivity

-Missed at early, treatable stage

12

Cost of low specificity

$
Possible hazards
Burden of tests, psych effects

13

Determinants of positive predictive value (2)

-Specificity of test
-Prev of disease

14

Higher positive predictive value when...

-Higher specificity and prev of disease

15

Bowlin et al: Behavioural Risk Factor Survey

-Poor validity/greatly underestimated prev: hypertension, hypercholesterolemia
-Fair validity, moderately underestimated prev: smoking, obesity
-No conclusions possible: diabetes

16

Prummel et al: breast cancer screening

-CR: lower sensitivity, less effective
-DR better than CR (so ON swapped devices)

17

Top 4 cancer sites

-Prostate, lung, breast, colon/rectum
-Prime targets for screening