rprev (version 1.0.5)

test_homogeneity: Inspects disease incidence for its compatibility with a homogeneous Poisson process.

Description

Calculates incidence by year of the registry data, along with mean incidence with confidence intervals. A smoothed cumulative incidence function is fit to the data for inspecting deviations in the registry data from a homogeneous Poisson process.

Usage

test_homogeneity(
  entry,
  year_start = "01-01",
  truncate_start = FALSE,
  truncate_end = FALSE,
  population_size = NULL,
  df = 4,
  proportion = 1e+05,
  level = 0.95,
  precision = 2
)

Arguments

entry

Vector of diagnosis dates for each patient in the registry in the format YYYY-MM-DD.

year_start

Date which to use to delimit years in the format MM-DD. See details for how this is used.

truncate_start

See details.

truncate_end

See details.

population_size

The population of the area covered by the registry. If not provided then only absolute incidence can be calculated.

df

The desired degrees of freedom of the smooth.

proportion

The denominator of the incidence rate.

level

The desired confidence interval width.

precision

The number of decimal places required.

Value

An S3 object of class incidence with the following attributes:

yearly_incidence

Vector of absolute incidence values for each included year of the registry

ordered_diagnoses

Vector of times (days) between diagnosis date and the earliest date of inclusion in the registry, ordered shortest to longest.

smooth

Smooth fitted to the cumulative incidence data.

index_dates

Dates delimiting the years in which incidence is calculated.

mean

List containing absolute yearly incidence as well as relative rates.

pvals

p-values resulting to a test of over and under dispersion on the incidence data respectively. Used to test the suitability of the homogeneous Poission process assumption.

dof

Degrees of freedom of the smooth.

Details

Annual incidence rates are calculated for every year that is present in entry, with years being delimited by the date specified in year_start that include every incident case. For example, under the default values, if the earliest incident date in entry is 1981-04-28, and the latest is 2016-12-16, then annual incidence rates will be calculated with the boundaries [1981-01-01, 1982-01-01), ..., [2016-01-01, 2017-01-01).

If year_start was specified as '09-01' then the boundaries would be [1980-09-01, 1981-09-01), ..., [2016-09-01, 2017-09-01).

The truncate_start and truncate_end arguments remove incident cases in the first and last years before and after the yearly boundaries respectively.

So if they were both TRUE, with year_start as '09-01' as before, then the boundaries would be [1981-09-01, 1982-09-01), ..., [2015-09-01, 2016-09-01), i.e. the incident cases in [1981-04-28, 1981-09-01) are discarded by truncate_start and those in [2016-09-01, 2016-12-16] removed by truncate_end.

This helps to ensure that annual incidence is measured on a time-scale appropriate for your registry.

Examples

# NOT RUN {
data(prevsim)

# }
# NOT RUN {
test_homogeneity(prevsim$entrydate)
# }
# NOT RUN {
# }