Table 2.

Quantitative bias analysis for outcome misclassification using positive predictive values and negative predictive values

Exposure-outcome cell counts observed in electronic health data source, not accounting for outcome misclassification
	Observed to have the outcome in data source	Observed to not have the outcome in data source
Exposed	$A$	$B$
Unexposed	$C$	$D$
Formula 1: Calculation of relative risk observed in electronic data in an analysis of electronic data, without consideration of outcome misclassification
RR(Observed) = $\frac{\frac{A}{A+B}}{\frac{C}{C+D}}$
Corrected distribution of outcomes by exposure using positive and negative predictive values 20
	With Outcome	Without Outcome
Exposed	$A^{\mathrm{*}}\mathrm{}=\mathrm{}A\left({\mathrm{PPV}}_1\right)\mathrm{}+B$(1-NPV1)	$B^{\mathrm{*}}=\mathrm{}(A+B)\mathrm{}–\left(A\right({\mathrm{PPV}}_1)\mathrm{}+\mathrm{}B$(1-NPV1))
Unexposed	$C^{\mathrm{*}}=C\left({\mathrm{PPV}}_0\right)\mathrm{}+\mathrm{}D$(1-NPV0)	$D^{\mathrm{*}}=\mathrm{}(C+D)\mathrm{}–\left(C\right({\mathrm{PPV}}_0)\mathrm{}+\mathrm{}D$(1-NPV0))
Formula 2: Calculation of corrected relative risk following corrected distribution of outcomes by exposure
RR(Corrected) = $\frac{\frac{A^{\mathrm{*}}}{A^{\mathrm{*}}+B^{\mathrm{*}}}}{\frac{C^{\mathrm{*}}}{C^{\mathrm{*}}+D^{\mathrm{*}}}}$

NPV₀: negative predictive value among unexposed; NPV₁: negative predictive value among exposed; PPV₀: positive predictive value among unexposed; PPV₁: positive predictive value among exposed; RR: relative risk.