Bayesian negative-binomial-family-based multistate Markov model for the evaluation of periodic population-based cancer screening considering incomplete information and measurement errors

Chen-Yang Hsu, Ming-Fang Yen, Anssi Auvinen, Sherry Chiu, Hsiu-Hsi Chen

Research output: Contribution to journalArticle

Abstract

Population-based cancer screening is often asked but hardly addressed by a question: “How many rounds of screening are required before identifying a cancer of interest staying in the pre-clinical detectable phase (PCDP)?” and also a similar one related to the number of screens required for stopping screening for the low risk group. It can be answered by using longitudinal follow-up data on repeated rounds of screen, namely periodic screen, but such kind of data are rather complicated and fraught with intractable statistical properties including correlated multistate outcomes, unobserved and incomplete (censoring or truncation) information, and imperfect measurements. We therefore developed a negative-binomial-family-based discrete-time stochastic process, taking sensitivity and specificity into account, to accommodate these thorny issues. The estimation of parameters was implemented with Bayesian Markov Chain Monte Carlo method. We demonstrated how to apply this proposed negative-binomial-family-based model to the empirical data similar to the Finnish breast cancer screening program.
Original languageEnglish
Pages (from-to)1-21
JournalStatistical Methods in Medical Research
DOIs
Publication statusPublished - Dec 15 2016

Fingerprint

Multi-state Model
Negative Binomial
Incomplete Information
Early Detection of Cancer
Measurement Error
Markov Model
Screening
Cancer
Stochastic Processes
Monte Carlo Method
Markov Chains
Evaluation
Population
Breast Neoplasms
Sensitivity and Specificity
Multi-state
Markov Chain Monte Carlo Methods
Censoring
Breast Cancer
Truncation

Cite this

@article{2068b6942e4c4eb390a7766b57c2ceca,
title = "Bayesian negative-binomial-family-based multistate Markov model for the evaluation of periodic population-based cancer screening considering incomplete information and measurement errors",
abstract = "Population-based cancer screening is often asked but hardly addressed by a question: “How many rounds of screening are required before identifying a cancer of interest staying in the pre-clinical detectable phase (PCDP)?” and also a similar one related to the number of screens required for stopping screening for the low risk group. It can be answered by using longitudinal follow-up data on repeated rounds of screen, namely periodic screen, but such kind of data are rather complicated and fraught with intractable statistical properties including correlated multistate outcomes, unobserved and incomplete (censoring or truncation) information, and imperfect measurements. We therefore developed a negative-binomial-family-based discrete-time stochastic process, taking sensitivity and specificity into account, to accommodate these thorny issues. The estimation of parameters was implemented with Bayesian Markov Chain Monte Carlo method. We demonstrated how to apply this proposed negative-binomial-family-based model to the empirical data similar to the Finnish breast cancer screening program.",
author = "Chen-Yang Hsu and Ming-Fang Yen and Anssi Auvinen and Sherry Chiu and Hsiu-Hsi Chen",
year = "2016",
month = "12",
day = "15",
doi = "10.1177/0962280216682284",
language = "English",
pages = "1--21",
journal = "Statistical Methods in Medical Research",
issn = "0962-2802",
publisher = "SAGE Publications Ltd",

}

TY - JOUR

T1 - Bayesian negative-binomial-family-based multistate Markov model for the evaluation of periodic population-based cancer screening considering incomplete information and measurement errors

AU - Hsu, Chen-Yang

AU - Yen, Ming-Fang

AU - Auvinen, Anssi

AU - Chiu, Sherry

AU - Chen, Hsiu-Hsi

PY - 2016/12/15

Y1 - 2016/12/15

N2 - Population-based cancer screening is often asked but hardly addressed by a question: “How many rounds of screening are required before identifying a cancer of interest staying in the pre-clinical detectable phase (PCDP)?” and also a similar one related to the number of screens required for stopping screening for the low risk group. It can be answered by using longitudinal follow-up data on repeated rounds of screen, namely periodic screen, but such kind of data are rather complicated and fraught with intractable statistical properties including correlated multistate outcomes, unobserved and incomplete (censoring or truncation) information, and imperfect measurements. We therefore developed a negative-binomial-family-based discrete-time stochastic process, taking sensitivity and specificity into account, to accommodate these thorny issues. The estimation of parameters was implemented with Bayesian Markov Chain Monte Carlo method. We demonstrated how to apply this proposed negative-binomial-family-based model to the empirical data similar to the Finnish breast cancer screening program.

AB - Population-based cancer screening is often asked but hardly addressed by a question: “How many rounds of screening are required before identifying a cancer of interest staying in the pre-clinical detectable phase (PCDP)?” and also a similar one related to the number of screens required for stopping screening for the low risk group. It can be answered by using longitudinal follow-up data on repeated rounds of screen, namely periodic screen, but such kind of data are rather complicated and fraught with intractable statistical properties including correlated multistate outcomes, unobserved and incomplete (censoring or truncation) information, and imperfect measurements. We therefore developed a negative-binomial-family-based discrete-time stochastic process, taking sensitivity and specificity into account, to accommodate these thorny issues. The estimation of parameters was implemented with Bayesian Markov Chain Monte Carlo method. We demonstrated how to apply this proposed negative-binomial-family-based model to the empirical data similar to the Finnish breast cancer screening program.

U2 - 10.1177/0962280216682284

DO - 10.1177/0962280216682284

M3 - Article

SP - 1

EP - 21

JO - Statistical Methods in Medical Research

JF - Statistical Methods in Medical Research

SN - 0962-2802

ER -