inst/app/www/misclassification.md

In dhaine/episensr.app: Shiny app for episensr, basic sensitivity analysis for epidemiological results

output: html_document: mathjax:"http://cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML" runtime: shiny

It performs a simple sensitivity analysis for disease or exposure misclassification. Confidence interval for odds ratio is computed as in Chu et al. (2006), for exposure misclassification.

Estimate of association can be biased if subjects are incorrectly categorized with respect to their exposure status or outcome, i.e. exposed/diseased subjects can be classified as non-exposed/non-diseased and vice versa. Most studies have some degree of misclassification, as there's no perfect instruments to obtain data (sensitivity and/or specificity are not perfect). Random error can also cause misclassification (e.g. data entry, missing data).

There are two type of misclassification bias, differential (misclassification is different in the groups to be compared) and nondifferential (misclassification is the same across groups to be compared).

In the Analysis tab, provide:

1. The 2-by-2 table of exposure and outcome,
2. The type of misclassification: exposure misclassification (corrections using sensitivity and specificity: nondifferential and independent errors) or outcome misclassification,
3. The following bias parameters:
   • Sensitivity of exposure (for exposure misclassification) or outcome (for outcome misclassification) classification among those with the outcome (for exposure misclassification) or exposure (for outcome misclassification),
   • Sensitivity of exposure (or outcome) classification among those without the outcome (or exposure),
   • Specificity of exposure (or outcome) classification among those with the outcome (or exposure), and
   • Specificity of exposure (or outcome) classification among those without the outcome (or exposure).

The data for the example come from:

• Fink, A.K., Lash, T.L. A null association between smoking during pregnancy and breast cancer using Massachusetts registry data (United States). Cancer Causes Control 2003;14:497-503.

- Chu, H., Zhaojie, W., Cole, S.R., Greenland, S., Sensitivity analysis of misclassification: A graphical and a Bayesian approach, Annals of Epidemiology 2006;16:834-841.

dhaine/episensr.app documentation built on Feb. 12, 2020, 9:12 a.m.