Epidemiology: 5) Exposure - Measurement
Terms
undefined, object
copy deck
- Major problems with measurement
-
Information bias.
prospective studies - nondifferential
When E is measured after D - diffential - Evaluating measurement methods
-
biological methods.
subject interview.
records. - 6 Measuring factors
-
1)Accuracy and complete
2)Nondifferential
3)Response rate
4)Temporal tracking
5)Availability
6)Cost - Current exposure: Biological or physical measurement
-
often gold std.
Current:
Accuracy and complete ++
nondifferential ++
response rate -
temproal tracking +
availability +
cost - - past exposure: biological or physical measurement
-
Past
Accuracy and complete --
nondifferential --
response rate --
temproal tracking --
availability --
cost -- - Coefficient of variation
- variation expresses as % of the mean
- Interview
- most common method but is problematic for past exposures or exposure after disease
- most common problems of interviews
-
poor technique.
fail to cross-check with other data sources.
insensitive to counterincentives to accurate reporting - Current exposure: Interview
-
Current:
Accuracy and complete ?
nondifferential +
response rate +
temproal tracking +
availability ++
cost + - Past exposure: Interview
-
Past
Accuracy and complete -
nondifferential -
response rate +
temproal tracking --
availability +
cost + - Current exposure: records
-
Current:
Accuracy and complete ?
nondifferential ++
response rate ++
temproal tracking ++
availability -
cost ++ - Past exposure: records
-
Past:
Accuracy and complete ?
nondifferential ++
response rate ++
temporal tracking ++
availability -
cost ++ - Assessing measurment quality
-
Requires gold standard
1)Categorical: Sensitivity, Specificity, PPV
2)Product moment correlation coefficient (Pearson) - Indices of reproducibility
-
Categorical: Kappa
Continuous: Intra-class correlation - Effects of nondifferential measurement error
- Bias towards Null
- Effect of differential exposure misclassification
- Spurious diffences between E+ and E-
- Exposure misclassification in a cohort study
-
Baseline measurement.
Unmeasured changes in exposure during followup - Exposure misclassification in a case-control study
-
1)Recall failure (non-differential)
2)Differential recall between cases and controls - Minimizing exposure misclassification
-
1)Good technique: Standardized, structureed questions, repeate measures.
2)Use of records
3)Substudy estimate misclassification, mathematical correlation