Let’s return to our mortality graph to visualize the excess cohort mortality that will be simulated in the clinic population. The extra-deaths are defined by the area between the clinic mortality rate (which is subject to late-entry underreporting) and the gold-standard cohort mortality rate (shaded region).

The excess cohort mortality rate in each year of life is converted to the expected number of unrecorded deaths in the clinic-based patient population in each year from birth to 25 years (total simulated deaths=684), and each death is randomly assigned a date of birth within the year. These extra, simulated deaths are appended to the clinic-based patient dataset, and time from birth to death is presented by gender using standard Kaplan-Meier methodology. The precision of simulated Kaplan-Meier statistics is examined using bootstrapping (1), a technique in wide use in epidemiology (e.g. 2;3). One thousand bootstrap realisations of the ‘excess mortality rate’ dataset allows the calculation of 95% bias corrected and accelerated (BC_A) confidence limits for Kaplan-Meier estimates and for associated Greenwood confidence limits. Results are compared to the standard statistical adjustment.

References

(1) Efron B, Tibshirani RJ. An Introduction to the Bootstrap. New York: Chapman and Hall, 1993.

(2) Hendriks JC, Satten GA, van Ameijden EJ, van Druten HA, Coutinho RA, van Griensven GJ. The incubation period to AIDS in injecting drug users estimated from prevalent cohort data, accounting for death prior to an AIDS diagnosis. AIDS 1998; 12(12):1537-1544.

(3) Tsodikov A, Hasenclever D, Loeffler M. Regression with bounded outcome score: evaluation of power by bootstrap and simulation in a chronic myelogenous leukaemia clinical trial. Stat Med 1998; 17(17):1909-1922.