An Epidemiologic Note on Verification Bias: Implications for Estimation of Rates