Table 4.
Simulation Scenario | Simulated levels of outcome misclassification | Mean overall outcome PPV PPV1, PPV0 | Mean overall outcome NPV NPV1, NPV0 | Median RR (95% SI) |
||
---|---|---|---|---|---|---|
From analysis of electronic data with outcome misclassification present |
Following QBA using PPV1, PPV0, NPV1, NPV0 |
Following QBA using PPV1, PPV0, and assuming nondifferential sensitivity |
||||
Formula 1 | Formula 2 | Formula 3 | ||||
Nondifferential misclassification | ||||||
1 | SN = 95%, SP = 99% | 84.1% 84.1%, 84.1% | 99.7% 99.7%, 99.7% | 1.00 (0.90-1.11) 5.0% type I error | 1.00 (0.90-1.11) | 1.00 (0.90-1.12) |
2 | SN = 90%, SP = 98% | 71.5% 71.4%, 71.5% | 99.4% 99.4%, 99.4% | 1.00 (0.90-1.11) 4.6% type I error | 1.00 (0.90-1.11) | 1.00 (0.88-1.11) |
Differential specificity | ||||||
3 | SN1 = 95%; SN0 = 95% | 79.8% | 99.7% | 2.43 | 1.00 | 1.00 |
SP1 = 90%; SP0 = 99% | 34.6%, 84.1% | 99.7%, 99.7% | (2.28-2.59) 100% type I error | (0.90-1.11) | (0.89-1.11) | |
4 | SN1 = 95%; SN0 = 95% | 72.9% | 99.7% | 0.86 | 1.00 | 1.00 |
SP1 = 99%; SP0 = 98% | 84.1%, 72.5% | 99.7%, 99.7% | (0.78-0.95) 80.7% type I error | (0.90-1.11) | (0.89-1.11) | |
Differential sensitivity | ||||||
5 | SN1 = 70%; SN0 = 95% | 84.0% | 99.7% | 0.78 | 1.00 | 0.74 |
SP1 = 99%; SP0 = 99% | 79.4%, 84.1% | 98.3%, 99.7% | (0.69-0.87) 99.2% type I error | (0.90-1.11) | (0.64-0.83) | |
6 | SN1 = 50%; SN0 = 95% | 83.9% | 99.6% | 0.60 | 1.00 | 0.52 |
SP1 = 99%; SP0 = 99% | 73.5%, 84.1% | 97.3%, 99.7% | (0.52-0.68) 100% type I error | (0.90-1.11) | (0.44-0.61) | |
Differential specificity and sensitivity | ||||||
7 | SN1 = 70%; SN0 = 95% | 72.7% | 99.7% | 0.67 | 1.00 | 0.73 |
SP1 = 99%; SP0 = 98% | 79.6%, 72.6% | 98.3%, 99.7% | (0.60-0.75) 100% type I error | (0.90-1.11) | (0.64-0.83) | |
8 | SN1 = 90%; SN0 = 95% | 79.7% | 99.7% | 2.38 | 1.00 | 0.95 |
SP1 = 90%; SP0 = 99% | 33.4%, 84.1% | 99.4%, 99.7% | (2.23-2.55) 100% type I error | (0.90-1.11) | (0.84-1.06) |
The 95% SI represents the 2.5th and 97.5th percentiles of the RR from across the 1000 replications for each simulated scenario. The 95% SI for the simulated RR of 1.00 before applying any misclassification was 0.90-1.11.
PPV0: positive predictive value among unexposed; PPV1: positive predictive value among exposed; QBA: quantitative bias analysis; SI: simulation interval; SN: overall outcome sensitivity; SN0: outcome sensitivity among unexposed; SN1: outcome sensitivity among exposed; SP: overall outcome specificity; SP0: outcome specificity among unexposed; SP1: outcome specificity among exposed.