Abstract
The colorectal cancer test dissemination application simulates the age of first test for colorectal cancer for an individual based on the type of colorectal test/examination (home FOBT or endoscopy), the individual's sex, and year of birth. The application can also simulate the age at which an individual received their first colorectal cancer test (either an FOBT or an endoscopy) and which type of test was received.
Cross-sectional survey data from the NHIS and BRFSS surveys were combined using a marginalized mixed-influence diffusion model to estimate the cumulative distribution for the time to first FOBT and endoscopy (Miglioretti and Brown, In Review). The age at first test prior to 2003 given birthyear and gender is then simulated based on this model. Note that the simulation does not model diagnosis of cancer or death, so the user of the simulation must truncate the age at first test at the time of diagnosis or death. The age at first test is truncated at the year 2003 to limit extrapolation beyond the last year of available data on colorectal cancer test use. Also note both surveys do not make a distinction between screening and diagnostic tests, so the model results should be interpreted as first test of any kind.
Summary
Description of simulation
Cross-sectional survey data from the NHIS and BRFSS surveys were combined using a marginalized mixed-influence diffusion model to estimate the cumulative distribution for the time to first FOBT and endoscopy (Miglioretti and Brown, In Review). The age at first test prior to 2003 given birth year and gender is then simulated based on this model. Note that the simulation does not model diagnosis of cancer or death, so the user of the simulation must truncate the age at first test at the time of diagnosis or death. The age at first test is truncated at the year 2003 to limit extrapolation beyond the last year of available data on colorectal cancer test use. Also note both surveys do not make a distinction between screening and diagnostic tests, so the model results should be interpreted as first test of any kind.
Estimated cumulative distribution for the time to first endoscopy. The estimated curve is a marginalized mixed-influence diffusion model of data from two different surveys. The letters represent the weighted proportion of individuals who said they had ever used/had the colorectal test/examination from the BRFSS (b or B) and the NHIS (N). For the BRFSS, in some years, only some states asked about colorectal cancer test use. In these years, the mean response will not represent the national average. These years are indicated with a lower-case b to distinguish them from years in which all states asked the question. Key: b=BRFFS (only some states asked about test use), B=BRFSS (all states asked about test use) N=NHIS (all states asked about test use)
The application uses a national level model to calculate the cumulative probability distribution that a person will receive either an FOBT or endoscopy test from the time the test was first introduced up to year 2003. This prior test-use probability is calculated based on the person's sex and year of birth.
Figure 1 depicts the cumulative distribution for males born in 1940. The simulated person is assigned a pseudo random number from a uniform distribution. If the person's random number is less than the 2003 testing probability (in the case of Figure 1 approximately 0.53, indicated by P) then the simulated person received the test sometime prior to 2003. The exact year of first test is established from the cumulative distribution by determining the year that corresponds to the individual's random number. For example, in figure 1, a random number of 0.25 corresponds to 1990 as the year of first test.
In order to determine which type of test (FOBT or endoscopy) was received first, the model assignes each person two correlated random values. One of the random values is used to calculate the age at which the person would receive their first FOBT, and the other random value is used to calculate the age at which they would receive their first endoscopy. In this model, endoscopy includes both sigmoidoscopy and colonoscopy. The age at testing calculations are performed using the national level model in the same manner as the individual age at first test calculation. Once an age at first FOBT and age at first endoscopy have been calculated for the individual, the test with the younger age is taken to be the first testing the person receives. If the two test ages are equal, the FOBT is the assigned test type.
Contact Information
If you have questions on the modeling used in the simulation or if you would like access to the simulation source code, please send an email to Diana Miglioretti miglioretti.d@ghc.org.
References
- Miglioretti DL, Brown ER. “A marginalized diffusion model for estimating age at first endoscopy examination from current-status data.” Applied Statistics (In Review)
Interface
Important: These models not designed for use by or for cancer patients or clinicians seeking treatment guidance for individual cases. Rather, they are largely based on retrospective surveillance data and have been built to aid our understanding of the impact of cancer control interventions (e.g., prevention, screening treatment) on population trends in incidence and mortality. Furthermore, the models here represent a small part of the larger CISNET effort and are used as population level inputs to cancer natural history and prognosis models. Please see the CISNET home page for more details.