Professional Documents
Culture Documents
Potential Error in Epidemiologic Study
Potential Error in Epidemiologic Study
Potential Error in Epidemiologic Study
EPIDEMIOLOGIC STUDY
EPIDEMIOLOGIC RESEARCH
BIAS
Exposure CONFOUNDING
A CHANCE
Exposure
B
Conclusion
Decision
OBJECTIVES
Recognize the possible research error
Understand the difference between
Random error
Bias
Confounding
Design?
Measurement?
Result?
NATURAL HISTORY OF THE DISEASE
Point of Onset of
exposure Symptoms
SCREENING PROCESS
Population
Screening
Re-test intervene
EXAMPLES OF SCREENING TESTS
Questions
ClinicalExaminations
Laboratory tests
Genetic tests
Radiologic examination
CONCLUSION OF EPIDEMIOLOGIC
RESEARCH
RESULT
Outcome Total
CAD (+) CAD (-)
Screening ECG + a b
ECG - c d
Total
Specificity
Sensitivity = a/a+c
SPECIFICITY
Outcome Total
CAD (+) CAD (-)
Screening ECG + a b
True Positive False Positive
ECG - c d
False Negative True Negative
Total
Proportion of individual who DO NOT have the disease
with negative screening result
The ability of the test to identify those WITHOUT
disease
Specificity = d/b+d
PREDICTIVE VALUE
Measure whether or not an individual, who
got a result of a screening test, actually has
a disease
Affected by
Specificity
Prevalence
sensitivity
POSITIVE PREDICTIVE VALUE
Outcome Total
CAD (+) CAD (-)
Screening ECG + a b
ECG - c d
Total
PPV = a/a+b
NEGATIVE PREDICTIVE VALUE
Outcome Total
CAD (+) CAD (-)
Screening ECG + a b
ECG - c d
Total
NPV = d/c+d
EXERCISE
Outcome Total
CAD (+) CAD (-)
Screening ECG + 36 15 51
ECG - 14 35 49
Total 50 50 100
Please calculate:
Sensitivity
Specificity
PositivePredictive Value
Negative Predictive Value
Internal validity
The degree to which the observed results of the
study are true
Inferences are correct regarding the participants in
the study
Internal validity
VALIDITY
SAMPLE SAMPLE
Selection bias
E+ E-
POPULATION
Measurement &
confounding bias
Chance
Out
External validity come
(generalizability)
CONCLUSION
External Validity
Generalizabilityof the result
Inferences are correct regarding the population
at risk
STUDY VALIDITY
AN APPROACH
The observed results IF:
(conclusion) occurred the role of chance is
because: small
Chance
bias can be reasonably
Random error
excluded
Bias
confounding is
Systematic error
Confounding addressed
Truth THEN
the study is internally
valid
VALIDITY, PERFORMANCE OF
MEASUREMENTS
Content validity:
Measurement includes all the dimension
Construct validity:
Measurement is related in a coherent way
Criterion validity:
Measurement predict a directly observable
phenomenon
RELIABILITY
Consistency of Measurement
Reproducibility over time
Consistency between different
coders/observers
Consistency among multiple indicators
Estimates of Reliability
Statisticalcoefficients that tell use how
consistently we measured something
FOUR ASPECTS OF RELIABILITY:
1. Stability
Consistency across time: repeat measurements
2. Reproducibility
Consistency between observer
3. Homogeneity
Consistency between different measures of the same
concept: use different items to get a conclusion of the same
concept
4. Accuracy
Lack of mistakes in measurement: good concept of
definition and procedures
Dedicated observers: training, motivation, concentration
RELATIONSHIP OF RELIABILITY TO
VALIDITY