surveil: Public health surveillance

The surveil R package provides time series models for routine public health surveillance tasks: model time trends in mortality or disease incidence rates to make inferences about levels of risk, cumulative and period percent change, age-standardized rates, and health inequalities.

surveil is an interface to Stan, a state-of-the-art platform for Bayesian inference. For analysis of spatial health data see the geostan R package.

Installation

surveil is available on CRAN; install from R using:

install.packages("surveil")

Vignettes

Review the package vignettes to get started:

• vignette("surveil-demo"): An introduction to public health modeling with surveil
• vignette("age-standardization"): How to age-standardize surveil model results
• vignette("measuring-inequality"): Assessing pairwise health differences with measures of inequality
• vignette("surveil-mcmc"): A brief introduction to Markov chain Monte Carlo (MCMC) with surveil

Also see the online documentation.

Usage

Model time series data of mortality or disease incidence by loading the surveil package into R together with disease surveillance data. Tables exported from CDC WONDER are automatically in the correct format.

library(surveil)
library(knitr)
data(cancer)

kable(head(cancer), 
      booktabs = TRUE,
      caption = "Table 1. A glimpse of cancer surveillance data")

Year	Age	Count	Population
1999	<1	866	3708753
1999	1-4	2959	14991152
1999	5-9	2226	20146188
1999	10-14	2447	19742631
1999	15-19	3875	19585857
1999	20-24	5969	18148795

Model trends in risk and easily view functions of risk estimates, such as cumulative percent change:

fit <- stan_rw(data = cancer,
               time = Year, 
               group = Age,
           cores = 4 # multi-core processing for speed
           )

fit_apc <- apc(fit)
plot(fit_apc, cumulative = TRUE)

*Cumulative percent change in US cancer incidence by age group*