Building an Endemic Channel with epiCo

1. What is an endemic channel?

Using an endemic channel is a visual strategy to represent the historical behavior of a disease in a specific region on an epidemic curve that defines the central tendency of the cases during a year and the upper and lower limits where cases are expected to vary. This plot provides three lines known as “\(\color{green}{\text{Safety}}\)”, “\(\color{orange}{\text{Warning}}\)”, and “\(\color{Red}{\text{Epidemic}}\)” bands (Figure 1), which are later used to define the epidemiological situation of the region based on the current (observed) cases.

The broader definition and methodology for the endemic channel were provided by Bortman (1999).

The data needed to build an endemic channel is the weekly or monthly incidence of the disease for previous years in the region of interest. epiCo asks for at least one year of data, but endemic channels are typically estimated with 5 to 7 years of information.

It is natural to presume that more years provide better estimations since the statistical analyses will be more robust, but it is important to have contextual information to ensure that transmission, surveillance, or demographic conditions (among other factors) have not changed during this period of time. Regarding the frequency of the data, weekly incidence may provide more useful information about the disease behavior and the moment in which an epidemiological warning should be raised; however, it is up to the experience of the users, and their context, whether this resolution of the data can be achieved.

Another common decision while building an endemic channel is to ignore previous epidemic years to avoid misrepresenting the traditional transmission dynamics. epiCo does not suggest epidemic years automatically, but its endemic_channel function does provide the option to annotate the outlier years and decide whether to include, ignore, or adjust them. More data-gathering tips and sources are provided later in this vignette.

2. Historical data needed to build an endemic channel

This section provides some strategies to obtain, handle, and gather the historical data needed to build an endemic channel. It presumes that the user has either the line list or incidence data for the disease of interest or the need to consult historical data from the Colombian National Surveillance System (SIVIGILA).

Regardless of the user´s case, the goal of this section is to create an incidence object with historical data on the disease. This object must account for the incidence of a single region (no groups are allowed in the endemic_channel function), it must have a weekly or monthly interval (epiweeks are also allowed), and it must represent at least one year of information.

To have a broader understanding of the incidence package, please refer to its vignettes.

library(epiCo)
library(incidence)

data("epi_data")
data_event <- epi_data

data_ibague <- data_event[data_event$cod_mun_o == 73001, ]

## Building the historic incidence data

incidence_historic <- incidence(data_ibague$fec_not,
  interval = "1 epiweek"
)

knitr::kable(incidence_historic[1:5, ])

3. Using epiCo’s endemic_channel function

After gathering the incidence data, users are ready to ask epiCo to build an endemic channel.

The endemic_channel function has the following parameters:

• incidence_historic: An incidence object with the case counts of the previous years

• observations: A numeric vector with the current observations (default = NULL)

• method: A string with the mean calculation method of preference (“median”, “mean”, or “geometric”) or “unusual_behavior” method (Poisson Distribution Test) (default = “geometric”)

• geometric_method: A string with the selected method for geometric mean calculation (see epiCo’s geometric_mean function) (default = “shifted”).

• outlier_years: A numeric vector with the outlier (epidemic) years (default = NULL).

• outliers_handling: A string with the handling decision regarding outlier years:

  • ignored = data from outlier years will not be taken into account (default).
  • included = data from outlier years will take into account.
  • replaced_by_median = data from outlier years will be replaced with the median and taken into account.
  • replaced_by_mean = data from outlier years will be replaced with the mean and taken into account.
  • replaced_by_geometric_mean = data from outlier years will be replaced with the geometric mean and taken into account.

• ci: A numeric value to specify the confidence interval to use with the geometric method (default = 0.95).

• plot: A boolean for displaying a plot (default = FALSE).

The output of epiCo’s endemic_channel function is a dataframe with the observation, historical mean, and confidence intervals, and a plot displaying the four epidemiological bands, the current observations (if provided), and the methods and outliers handling set by the user.

observations <- sample(5:40, 52, replace = TRUE)

outlier_years <- c(2016, 2019)

ibague_endemic_chanel <- endemic_channel(
  incidence_historic = incidence_historic,
  observations = observations,
  method = "geometric",
  outlier_years = outlier_years,
  plot = TRUE
)
#> Data after2022-12-25 were not used for the endemic channel calculation.

4. Interpretation and communication of results

The endemic channel is a visualization tool to stratify the epidemiological situation of a region. The instructions to gather, process and interpret the results are responsibility of the governmental institutions.

The communication of results is typically done in a framework of risk assessment, early warning protocols, and epidemic response leaded by public health officers. It is important to adjust the use of epiCo to the official guidelines from where is implemented.