Figure 3. The impact cause-of-death misclassification on the 10-year cumulative incidence of lung cancer deaths (LC-CID) among patients aged 60 to 74 with stage I disease.
In each scenario, a refers to the probability of under-reporting lung cancer deaths, and b refers to the probability of over-reporting lung cancer deaths. Results are presented for increments in b given increasing values of a (left panel) or increments in a given increasing values of b (right panel). Based on literature, the plausible values of under-reporting and over-reporting are a = 15% and b = 1%. The differences between observed 10-year LC-CIDs and corrected 10-year LC-CIDs were reported as biases in each scenario. A positive value of bias indicates that the observed 10-year LC-CID overestimated the corrected 10-year LC-CID; a negative value of bias indicates that the observed 10-year LC-CID underestimated the corrected 10-year LC-CID. The gray dashed line at y = 0 represents the ideal situation of no bias.