Estimates of disease-specific incidence, prevalence and mortality specified by age are important information factors for estimating the burden of disease. Publications on prevalence estimates from the population-based registry generally consider all people with a past diagnosis of cancer as prevalent cases without taking into account the possibility of getting better. But today, many cancer patients are actually cured of the disease. It is therefore important to take recovery into account in the estimates of prevalence.
In this work three methods of estimating age-specific prevalence are presented, two of which allow us to estimate age-specific prevalence of non recovery. On the one hand, thanks to a four-state stochastic model and Poisson process, expressions of age-specific prevalence and non recovery prevalence can be built. From these expressions and using an actuarial method of estimation, this leads to the Transition Rate Method (TRM) (the first approach). Moreover, assuming the rare disease and using a parametric method of estimation, this leads to the Parametric Model (PM) (the second approach). On the other hand, the Counting Method (CM) is presented. Contrary to the other method, transition rate estimates are not required. The Counting Method counts all subjects who are known to have survived for a certain calendar time t and adds an estimate of the number of survivors among those who were alive or lost from follow-up before t.
In section 1, the concept of recovery is defined. The definitions of age-specific prevalence and non recovery prevalence are also outlined. One approach is to use data from disease registries to estimate various intensity functions. A second approach [CD97] is to use data from disease registries to estimate age-specific incidence and relative survival. In both approaches, prevalence and non recovery prevalence are thus determined. A third approach the counting method [GKMS99] [FKMN86], estimates the number of disease survivors in the population. These methods are described in section 2. In section 3, the application of these models is illustrated using data from the Surveillance, Epidemiology, and End Results [SEERD03] colorectal Cancer Registry in Connecticut. Colorectal cancer is the second most common cancer in developed countries.
Was this article helpful?