JMIR Publications

ABSTRACT

Background:

Monitoring disease incidence rates over time with population surveillance data is fundamental to public health research and practice. Bayesian disease monitoring methods provide advantages over conventional methods including greater flexibility in model specification and the ability to conduct formal inference on model-derived quantities of interest. However, software platforms for Bayesian inference are often inaccessible to non-specialists.

Objective:

This paper introduces a Bayesian methodology and open-source software package, surveil, for time series modeling of disease incidence and mortality. Given case count and population-at-risk data, the software enables health researchers to draw inferences about underlying risk and derivative quantities like age-standardized rates, annual and cumulative percent change, and measures of inequality.

Methods:

We specify a Poisson likelihood for case counts and model trends in log-risk using a first-difference (random-walk) prior. We also extend the model for correlated time series, such as observations stratified by demographic group. Models in the surveil R package were built using the Stan platform for Bayesian inference. We demonstrate the methodology and software by analyzing colorectal cancer (CRC) incidence rates by race-ethnicity among non-Hispanic Black (Black), non-Hispanic White (White), and Hispanic adults ages 50-79 in each of the four largest metropolitan statistical areas (MSAs) of Texas, years 1999-2017. We measure cancer inequality using complementary measures of pairwise inequality and Theil’s index for nested population structures.

Results:

Our demonstration analysis highlights a large and persistent Black-White inequality in CRC incidence rates. The Black-White rate difference (RD) rose from 35 per 100,000 (95% CI 23-47) in 1999 to 45 (95% CI 38-54) by 2008, and then fell to about 31 (95% CI 24-38) by 2017. The Black-White disparity in these MSAs accounts cumulatively for about 3,206 (95% CI 2,976-3,435) CRC cases, or 26% (95% CI 25%-28%) of all Black CRC incidence over this period. By Theil's index, no reduction of racial-ethnic inequality was achieved during this period.

Conclusions:

Our methodology and software can help public health agencies measure health inequalities and evaluate progress towards health equity goals. Advantages of this method over current common practice include: 1) no piecewise linearity constraints on the model space; and 2) the ability to make formal probability statements on any model-derived quantity of interest. Our analysis of urban CRC incidence indicates that excess Black CRC risk warrants special attention as a cancer prevention priority in Texas.