Package 'expard' reference manual

Title:	Drug EXPosures and ADRs
Description:	An R package for fitting complex drug exposure and adverse drug reaction (ADR) relationships
Authors:	Louis Dijkstra
Maintainer:	Louis Dijkstra <[email protected]>
License:	GPL-3
Version:	1.0.0
Built:	2025-01-04 05:56:51 UTC
Source:	https://github.com/bips-hb/expard

Apply Function to Subset Drug History

Description

Observations might not be present for certain patients, since they entered the data set later time = 1 or before the end. This function applies, for example, a risk model only to the observed time points.

Usage

apply_function_to_observed_timepoints(
  drug_history,
  fn = expard::risk_model_current_use()
)
apply_function_to_observed_timepoints(
  drug_history,
  fn = expard::risk_model_current_use()
)

Arguments

`drug_history`	The drug history. Unobserved time points are denoted by `NA`
`fn`	A function (Default `risk_model_current_use`).

Value

Vector of the same length as drug_history. None observed time points are denoted by NA

Creating a $2 \times 2$ Table

Description

Creates a $2 \times 2$ contingency table for a drug and ADR.

A table is structured in the following form:

	ADR	not ADR
drug	`a`	`c`
not drug	`b`	`d`

Usage

create2x2table(drug_ADR_pair, method = c("time-point", "drug-era", "patient"))
create2x2table(drug_ADR_pair, method = c("time-point", "drug-era", "patient"))

Arguments

`drug_ADR_pair`	A drug-ADR pair; see `generate_cohort`
`method`	Method used to construct the table; either `time-point`, `drug-era` and `patient`. See the description for more information (Default: `time-point`)

Value

A cont_table object; a list with

`a`, `b`, `c`, `d`	The counts in the table
`method`	Method that was used to construct the table
`n`	Total count (`n = a + b + c + d`). The interpretation depends on the `method` used

Ways to construct the table

The counts can be constructed in three different ways. See for two of them Zorych et al. (2013).

By individual time-points (`time-point`)

Each time-point is counted separately. The counts are the number of time points that

a - the drug was prescribed and the ADR occurred
b - the drug was not prescribed but the ADR occurred
c - the drug was prescribed but the ADR did not occur
d - the drug was not prescribed and the ADR did not occur

Note that in this case the total count n = a + b + c + d is the same as the total number of time points observed, i.e., the total number of patients times the number of time points observed: n_patients * simulation_time.

By individual patients (`patient`)

In this case, individual patients are counted. The counts are the number of patients that

a - were prescribed the drug and did experience the ADR
b - were never prescribed the drug and did experience the ADR
c - were prescribed the drug and never experienced the ADR
d - were never prescribed the drug and never experienced the ADR

In this case, the total count n = a + b + c +d is the same as the number of patients, n_patients.

By individual patients (`drug-era`)

In this case we look at drug-eras, i.e., periods in which the patients was prescribed or not prescribed the drug for a longer time. For example, if the patient was prescribed the drug from time point 3 to 6, then that period is called a drug-era. The counts are the number of drug- and non-drug eras in which

a - the drug was prescribed and the ADR occurred
b - the drug was not prescribed but the ADR occurred
c - the drug was prescribed but the ADR did not occur
d - the drug was not prescribed and the ADR did not occur

In this case, the total count n is the total number of drug- and non-drug eras.

References

Zorych, I., Madigan, D., Ryan, P., & Bate, A. (2013). Disproportionality methods for pharmacovigilance in longitudinal observational databases. Statistical Methods in Medical Research, 22(1), 39–56. https://doi.org/10.1177/0962280211403602

Examples

set.seed(1)
cohort <- generate_cohort(n_patients = 200) 

# create the 2x2 contingency table per time-point, 
# drug-era and patient: 
create2x2table(cohort, method = "time-point")
create2x2table(cohort, method = "drug-era")
create2x2table(cohort, method = "patient")
set.seed(1)
cohort <- generate_cohort(n_patients = 200) 

# create the 2x2 contingency table per time-point, 
# drug-era and patient: 
create2x2table(cohort, method = "time-point")
create2x2table(cohort, method = "drug-era")
create2x2table(cohort, method = "patient")

$2 \times 2$ Tables

Description

Creates the $2 \times 2$ contingency tables for all drug-ADR pairs in a specific cohort. The function is basically a wrapper for create2x2table

Usage

create2x2tables(
  cohort,
  method = c("time-point", "drug-era", "patient"),
  verbose = TRUE
)
create2x2tables(
  cohort,
  method = c("time-point", "drug-era", "patient"),
  verbose = TRUE
)

Arguments

`cohort`	A cohort; see `generate_cohort`
`method`	Method used to construct the table; either `time-point`, `drug-era` and `patient`. See the description for more information (Default: `time-point`)

Details

A table is structured in the following form:

	ADR	not ADR
drug	`a`	`c`
not drug	`b`	`d`

Value

A list of cont_table objects

Ways to construct the table

The counts can be constructed in three different ways. See for two of them Zorych et al. (2013).

By individual time-points (`time-point`)

Each time-point is counted separately. The counts are the number of time points that

a - the drug was prescribed and the ADR occurred
b - the drug was not prescribed but the ADR occurred
c - the drug was prescribed but the ADR did not occur
d - the drug was not prescribed and the ADR did not occur

By individual patients (`patient`)

In this case, individual patients are counted. The counts are the number of patients that

a - were prescribed the drug and did experience the ADR
b - were never prescribed the drug and did experience the ADR
c - were prescribed the drug and never experienced the ADR
d - were never prescribed the drug and never experienced the ADR

In this case, the total count n = a + b + c +d is the same as the number of patients, n_patients.

By individual patients (`drug-era`)

a - the drug was prescribed and the ADR occurred
b - the drug was not prescribed but the ADR occurred
c - the drug was prescribed but the ADR did not occur
d - the drug was not prescribed and the ADR did not occur

In this case, the total count n is the total number of drug- and non-drug eras.

References

Examples

set.seed(1)
cohort <- generate_cohort(n_patients = 200) 

# create the 2x2 contingency table per time-point, 
# drug-era and patient: 
create2x2table(cohort, method = "time-point")
create2x2table(cohort, method = "drug-era")
create2x2table(cohort, method = "patient")
set.seed(1)
cohort <- generate_cohort(n_patients = 200) 

# create the 2x2 contingency table per time-point, 
# drug-era and patient: 
create2x2table(cohort, method = "time-point")
create2x2table(cohort, method = "drug-era")
create2x2table(cohort, method = "patient")

First Prescription

Description

determine_first_perscription returns a time point for the first prescription of the drug of interest given the time frame of the patient (simulation_time) and the chance for it be prescribed is min_chance_drug. The patient, in this case, is always prescribed the drug at least once. The time point follows a $\Gamma$ distribution.

Usage

determine_first_prescription(n_patients, simulation_time, min_chance_drug)
determine_first_prescription(n_patients, simulation_time, min_chance_drug)

Arguments

`n_patients`	Number of patients (Default: 1)
`simulation_time`	The total number of time points for this patient
`min_chance_drug`	The probability of the drug being prescribed at any given time point

Details

Note: we assume here that the probability of the drug being prescribed does not depend on any previous prescriptions.

Value

A time point between 1 and simulation_time

Examples

set.seed(1)
determine_first_prescription(n_patients = 4, simulation_time = 10, min_chance_drug = 0.1)
# -> [1] 2 3 5 9 
set.seed(1)
determine_first_prescription(n_patients = 4, simulation_time = 10, min_chance_drug = 0.1)
# -> [1] 2 3 5 9

Compute Frequency Table

Description

Determine how often different values from time_steps_ago appear in the data set and how often the ADR occurs for each of them. This is used for speeding up the computation of the loglikeihood

Usage

determine_frequency_unique_values(mat, adr_history)
determine_frequency_unique_values(mat, adr_history)

Arguments

`mat`	Matrix with relevant values (depends on model)
`adr_history`	History of ADRs

Value

Tibble

`expard`: A package for simulating electronic health care data and fitting models

Author(s)

Louis Dijkstra

Fit a Model to Cohort Data

Description

fit_model fits a specified risk model risk_model to cohort data.

Usage

fit_model(
  pair,
  model = c("no-association", "current-use", "past-use", "withdrawal", "delayed",
    "decaying", "delayed+decaying", "long-term"),
  zero_patients = 0,
  zero_timepoints = 0,
  method = c("L-BFGS-B", "Nelder-Mead", "BFGS", "CG", "SANN", "Brent"),
  maxiter = 1000,
  parameters = list(),
  mc.cores = 1
)
fit_model(
  pair,
  model = c("no-association", "current-use", "past-use", "withdrawal", "delayed",
    "decaying", "delayed+decaying", "long-term"),
  zero_patients = 0,
  zero_timepoints = 0,
  method = c("L-BFGS-B", "Nelder-Mead", "BFGS", "CG", "SANN", "Brent"),
  maxiter = 1000,
  parameters = list(),
  mc.cores = 1
)

Arguments

`model`	Label for a risk model. Can be either ...
`method`	Methods used by the base `optim`-solver
`maxiter`	Maximum number iterations for the `optim`-solver
`cohort`	A cohort dataset. See details below.
`start`	Starting point for the base `optim`-solver (Default: `c(-1,1)`)
`control`	List used by the base `optim`-solver

Value

A model fit. A list with the items

`n_patients`	Total number of patients in the dataset
`simulation_time`	Total number of time points
`loglikelihood`	The maximal log-likelihood value
`beta0`	Intercept for the logistic model
`beta`	Parameter for the logistic model
`prob_no_adr_with_drug`	Estimated probability of no ADR occurring
`prob_ADR_with_drug`	Estimated probability of ADR occurring
`convergence`	Convergence value of `optim`. `0` means that the algorithm converged

Patient models

fit_model can currently only deal with the patient_model_uninformative. More complex patient models, such as patient_model_sex require the estimation of more parameters which is currently not supported.

Cohort data object

Minimal requirement is that the cohort is a list with two matrix with the items

drug_history A binary matrix of size n_patients x simulation_time describing the drug prescriptions.
adr_history A binary matrix of size n_patients x simulation_time describing ADR histories

See check_cohort and generate_cohort for more details on the cohort object.

Examples

cohort <- expard::generate_cohort(n_patients = 1000, 
                                  simulation_time = 100, 
                                  risk_model = expard::risk_model_immediate(), 
                                  verbose = TRUE, 
                                  min_chance_drug = probability_model_constant(.3), 
                                  max_chance_drug = probability_model_constant(.7), 
                                  min_chance_adr = probability_model_constant(.3),
                                  max_chance_adr = probability_model_constant(.6))
# fit the no effect model
fit_model(cohort, risk_model = expard::risk_model_no_effect()) 

# fit the true immediate effect model
fit_model(cohort, risk_model = expard::risk_model_immediate())
# note that the estimators are close to the truth (.3 and .6) 
cohort <- expard::generate_cohort(n_patients = 1000, 
                                  simulation_time = 100, 
                                  risk_model = expard::risk_model_immediate(), 
                                  verbose = TRUE, 
                                  min_chance_drug = probability_model_constant(.3), 
                                  max_chance_drug = probability_model_constant(.7), 
                                  min_chance_adr = probability_model_constant(.3),
                                  max_chance_adr = probability_model_constant(.6))
# fit the no effect model
fit_model(cohort, risk_model = expard::risk_model_no_effect()) 

# fit the true immediate effect model
fit_model(cohort, risk_model = expard::risk_model_immediate())
# note that the estimators are close to the truth (.3 and .6)

Returns ADR History

Description

Generates a binary vector of ADR occurrences given a drug history and a risk model.

Usage

generate_adr_history(drug_history, risk_model, min_chance, max_chance, ...)
generate_adr_history(drug_history, risk_model, min_chance, max_chance, ...)

Arguments

`drug_history`	Binary vector denoting the drug prescriptions history
`risk_model`	A risk model, e.g., `risk_model_current_use`
`min_chance`	The probability of the ADR when the drug history has no effect
`max_chance`	The probability of the ADR when the drug history has the highest possible effect

Value

Binary vector of the same length as drug_history

Generate a Cohort

Description

Generates a cohort which is a collection of drug-ADR pair. It is a wrapper function for the generate_patient function.

Usage

generate_cohort(
  n_patients = 100,
  simulation_time = 100,
  n_drug_ADR_pairs = 50,
  risk_model = rep("risk_model_current_use()", n_drug_ADR_pairs),
  min_chance_drug = rep(0.1, n_drug_ADR_pairs),
  avg_duration = rep(5, n_drug_ADR_pairs),
  max_chance_drug = rep(NULL, n_drug_ADR_pairs),
  prob_guaranteed_exposed = rep(1, n_drug_ADR_pairs),
  min_chance = rep(0.1, n_drug_ADR_pairs),
  max_chance = rep(0.4, n_drug_ADR_pairs),
  verbose = FALSE
)
generate_cohort(
  n_patients = 100,
  simulation_time = 100,
  n_drug_ADR_pairs = 50,
  risk_model = rep("risk_model_current_use()", n_drug_ADR_pairs),
  min_chance_drug = rep(0.1, n_drug_ADR_pairs),
  avg_duration = rep(5, n_drug_ADR_pairs),
  max_chance_drug = rep(NULL, n_drug_ADR_pairs),
  prob_guaranteed_exposed = rep(1, n_drug_ADR_pairs),
  min_chance = rep(0.1, n_drug_ADR_pairs),
  max_chance = rep(0.4, n_drug_ADR_pairs),
  verbose = FALSE
)

Arguments

`n_patients`	Number of patients
`simulation_time`	The total number of time steps
`n_drug_ADR_pairs`	Number of drug-ADR pairs simulated
`risk_model`	Vector with risk models. Each risk model is given as a string, e.g., `"risk_model_current_use()"`. The vector must have the length `n_drug_ADR_pairs`#'
`min_chance_drug`	Vector with the probabilities of the drug being prescribed when the drug history and the ADR history have no effect. Must have a length of `n_drug_ADR_pairs`.
`avg_duration`	Average number of time points a patient is exposed once exposed to the drug (Determines `max_chance`) Must have a length of `n_drug_ADR_pairs`.
`max_chance_drug`	The probability of the ADR when the drug history and the ADR history have the highest possible effect (Default: `NULL`) Must have a length of `n_drug_ADR_pairs`.
`min_chance`	The probability of the ADR when the drug history has no effect Must have a length of `n_drug_ADR_pairs`.
`max_chance`	The probability of the ADR when the drug history has the highest possible effect Must have a length of `n_drug_ADR_pairs`.
`verbose`	Show progress bar (Default: `FALSE`)

Value

A cohort object; A list of length n_drug_ADR_pairs. Each entry contains two binary matrices; one for the drug histories of all patients and one for the ADR histories of all patients. Each row is a patient

Generate a Drug-ADR pair

Description

Generates the drug- and ADR history

Usage

generate_drug_ADR_pair(
  simulation_time = 100,
  risk_model = risk_model_current_use(),
  min_chance_drug = 0.1,
  avg_duration = 5,
  max_chance_drug = NULL,
  guaranteed_exposed = TRUE,
  min_chance = 0.1,
  max_chance = 0.4
)
generate_drug_ADR_pair(
  simulation_time = 100,
  risk_model = risk_model_current_use(),
  min_chance_drug = 0.1,
  avg_duration = 5,
  max_chance_drug = NULL,
  guaranteed_exposed = TRUE,
  min_chance = 0.1,
  max_chance = 0.4
)

Arguments

`simulation_time`	The total number of time steps
`risk_model`	One of the risk models
`min_chance_drug`	The probability of the drug being prescribed when the drug history and the ADR history have no effect
`avg_duration`	Average number of time points a patient is exposed once exposed to the drug (Determines `max_chance`)
`max_chance_drug`	The probability of the ADR when the drug history and the ADR history have the highest possible effect (Default: `NULL`)
`guaranteed_exposed`	If `TRUE`, the patient is exposed to the drug at least once
`min_chance`	The probability of the ADR when the drug history has no effect
`max_chance`	The probability of the ADR when the drug history has the highest possible effect

Value

A list with

`drug_history`	binary vector of the drug history
`adr_history`	binary vector of the ADR history

Examples

drug_ADR_pair <- expard::generate_drug_ADR_pair()
drug_ADR_pair <- expard::generate_drug_ADR_pair()

Generate Drug Prescription History

Description

Returns whether or not the drug is prescribed at the next time point (1) or not (0). The probability depends on the drug prescriptions, the ADR history and the drug_model.

Usage

generate_drug_history(
  simulation_time = 10,
  min_chance,
  avg_duration = 5,
  max_chance = NULL,
  guaranteed_exposed = TRUE,
  ...
)
generate_drug_history(
  simulation_time = 10,
  min_chance,
  avg_duration = 5,
  max_chance = NULL,
  guaranteed_exposed = TRUE,
  ...
)

Arguments

`simulation_time`	Total number of time points
`min_chance`	The probability of the drug being prescribed when the drug history and the ADR history have no effect
`avg_duration`	Average number of time points a patient is exposed once exposed to the drug (Determines `max_chance`)
`max_chance`	The probability of the ADR when the drug history and the ADR history have the highest possible effect (Default: `NULL`)
`guaranteed_exposed`	If `TRUE`, the patient is exposed to the drug at least once

Value

1 or 0

Examples

generate_drug_history(simulation_time = 10, 
                      min_chance = .1,
                      avg_duration = 5
                      guaranteed_exposed = TRUE)
generate_drug_history(simulation_time = 10, 
                      min_chance = .1,
                      avg_duration = 5
                      guaranteed_exposed = TRUE)

Generate a Patient

Description

Generates a patient, which is a collection of different drug-ADR pairs, see function generate_drug_ADR_pair.

Usage

generate_patient(
  simulation_time = 100,
  n_drug_ADR_pairs = 50,
  risk_model = rep("risk_model_current_use()", n_drug_ADR_pairs),
  min_chance_drug = rep(0.1, n_drug_ADR_pairs),
  avg_duration = rep(5, n_drug_ADR_pairs),
  max_chance_drug = rep(NULL, n_drug_ADR_pairs),
  guaranteed_exposed = rep(TRUE, n_drug_ADR_pairs),
  min_chance = rep(0.1, n_drug_ADR_pairs),
  max_chance = rep(0.4, n_drug_ADR_pairs)
)
generate_patient(
  simulation_time = 100,
  n_drug_ADR_pairs = 50,
  risk_model = rep("risk_model_current_use()", n_drug_ADR_pairs),
  min_chance_drug = rep(0.1, n_drug_ADR_pairs),
  avg_duration = rep(5, n_drug_ADR_pairs),
  max_chance_drug = rep(NULL, n_drug_ADR_pairs),
  guaranteed_exposed = rep(TRUE, n_drug_ADR_pairs),
  min_chance = rep(0.1, n_drug_ADR_pairs),
  max_chance = rep(0.4, n_drug_ADR_pairs)
)

Arguments

`simulation_time`	The total number of time steps
`n_drug_ADR_pairs`	Number of drug-ADR pairs simulated
`risk_model`	Vector with risk models. Each risk model is given as a string, e.g., `"risk_model_current_use()"`. The vector must have the length `n_drug_ADR_pairs`#'
`min_chance_drug`	Vector with the probabilities of the drug being prescribed when the drug history and the ADR history have no effect. Must have a length of `n_drug_ADR_pairs`.
`avg_duration`	Average number of time points a patient is exposed once exposed to the drug (Determines `max_chance`) Must have a length of `n_drug_ADR_pairs`.
`max_chance_drug`	The probability of the ADR when the drug history and the ADR history have the highest possible effect (Default: `NULL`) Must have a length of `n_drug_ADR_pairs`.
`guaranteed_exposed`	If `TRUE`, the patient is exposed to the drug at least once Must have a length of `n_drug_ADR_pairs`.
`min_chance`	The probability of the ADR when the drug history has no effect Must have a length of `n_drug_ADR_pairs`.
`max_chance`	The probability of the ADR when the drug history has the highest possible effect Must have a length of `n_drug_ADR_pairs`.

Value

A list of drug-ADR pairs, see generate_drug_ADR_pair

Hello, World!

Description

Prints 'Hello, world!'.

Usage

hello()
hello()

Examples

hello()
hello()

Negative of the Log-Likelihood Function

Description

loglikelihood returns the negative log-likelihood function, defined as

$\pi = \frac{\exp(\beta_0 + \code{risks}\cdot \beta)}{1 + \exp(\beta_0 + \code{risks}\cdot \beta)}$

where $\pi$ is the probability.

Usage

loglikelihood(betas, risks, adr_history)
loglikelihood(betas, risks, adr_history)

Arguments

`betas`	Vector with two entries: `beta0, beta`
`risks`	Risks determined by the risk model. The risk model was applied to the observed drug prescription history

Plot the Fit

Description

Plots the BIC for the different risk models that were fitted by the function fit_all_models.

Usage

plot_fit(
  fit,
  x_label = "model",
  title = "",
  y_range = NULL,
  past_values = NULL
)
plot_fit(
  fit,
  x_label = "model",
  title = "",
  y_range = NULL,
  past_values = NULL
)

Arguments

`fit`	The fit
`past_values`	If not `NULL`, a select number of values for past are used for the plot
`xlab`	The x-label

Value

ggplot

Plot Risk

Description

Plots the risk function's behavior for a given drug prescription history. The $x$ -axis denote the drug prescriptions and the $y$ -axis represent the risk level, i.e., 1 is maximal risk, 0 minimal risk.

Usage

plot_risk(
  drug_history = c(rep(0, 4), rep(1, 6), rep(0, 10)),
  risk_model = risk_model_immediate(),
  simulation_time = NULL,
  title = "",
  ylim = c(0, 1),
  shaded_area = TRUE,
  fill = "black",
  alpha = 0.3
)
plot_risk(
  drug_history = c(rep(0, 4), rep(1, 6), rep(0, 10)),
  risk_model = risk_model_immediate(),
  simulation_time = NULL,
  title = "",
  ylim = c(0, 1),
  shaded_area = TRUE,
  fill = "black",
  alpha = 0.3
)

Arguments

`drug_history`	Binary vector denoting the drug prescription history. (Default: `0 0 0 0 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0`)
`risk_model`	One of the risk models
`simulation_time`	If not `NULL`, total number of time points, i.e., `0 0 0 0 1 1 1 1 1 1 0 0 0 0 ...`. (Default: `NULL`)
`title`	Title of the plot
`ylim`	The limits of the y-axis: `c(ymin, ymax)` (Default: `c(0,1)`)
`shaded_area`	If `TRUE`, periods in which drugs were prescribed are shaded in a gray tone. Otherwise, the drug prescription periods are denoted by vertical dashed red lines (Default: `TRUE`)
`fill`	Color used for the shaded area (Default: `black`)
`alpha`	Alpha value for shaded area (Default: `.3`)

Value

A ggplot2 plot

Examples

drug_history <- c(rep(0, 4), rep(1, 6), rep(0, 10))
risk_model <- risk_model_immediate() 

# create the plot
p <- plot_risk(drug_history, 
               risk_model, 
               title = "Direct effect model")
p
drug_history <- c(rep(0, 4), rep(1, 6), rep(0, 10))
risk_model <- risk_model_immediate() 

# create the plot
p <- plot_risk(drug_history, 
               risk_model, 
               title = "Direct effect model")
p

Print function for `generate_cohort`

Description

Print function for generate_cohort

Usage

## S3 method for class 'cohort'
print(cohort)
## S3 method for class 'cohort'
print(cohort)

Print function for 2x2 tables

Description

Print function for 2x2 tables

Usage

## S3 method for class 'cont_table'
print(table)
## S3 method for class 'cont_table'
print(table)

Function for printing a drug-ADR pair

Description

Function for printing a drug-ADR pair

Usage

## S3 method for class 'drug_ADR_pair'
print(drug_ADR_pair)
## S3 method for class 'drug_ADR_pair'
print(drug_ADR_pair)

Print function for the hccd fit_model

Description

Print function for the hccd fit_model

Usage

## S3 method for class 'expardfit2'
print(fit)
## S3 method for class 'expardfit2'
print(fit)

Function for printing a patient

Description

Function for printing a patient

Usage

## S3 method for class 'patient'
print(patient)
## S3 method for class 'patient'
print(patient)

Risk Model 'Immediate'

Description

Usage

risk_model_current_use()
risk_model_current_use()

Value

A risk model

Examples

drug_history <- c(1, 0, 1, 0, 0)
risk_model <- risk_model_immediate() 
risk_model(drug_history) 
drug_history <- c(1, 0, 1, 0, 0)
risk_model <- risk_model_immediate() 
risk_model(drug_history)

Risk Model 'Decaying'

Description

Risk Model 'Decaying'

Usage

risk_model_decaying(rate)
risk_model_decaying(rate)

Arguments

rate

The rate with which the risk dissipates

Value

A risk model

Examples

drug_history <- c(1, 0, 1, 0, 0)

risk_model <- risk_model_withdrawal(rate = 1.2) 
risk_model(drug_history) 
drug_history <- c(1, 0, 1, 0, 0)

risk_model <- risk_model_withdrawal(rate = 1.2) 
risk_model(drug_history)

Risk Model 'Delayed'

Description

Risk Model 'Delayed'

Usage

risk_model_delayed(mu, sigma)
risk_model_delayed(mu, sigma)

Arguments

rate

The rate with which the risk dissipates

Value

A risk model

Examples

drug_history <- c(1, 0, 1, 0, 0)

risk_model <- risk_model_withdrawal(rate = 1.2) 
risk_model(drug_history) 
drug_history <- c(1, 0, 1, 0, 0)

risk_model <- risk_model_withdrawal(rate = 1.2) 
risk_model(drug_history)

Risk Model 'Withdrawal'

Description

A risk model reflects how the probability of suffering the ADR changes with the drug exposures of a patient. It returns 0 when the drug prescription history has no effect, and 1 when the patient is at maximal risk. In this case, once the patient is no longer exposed to the drug, the probability of the ADR peaks and then dissipates exponentially. Let $\tau$ be the number of time points since the patient was prescribed the drug last. The return value is the given by

$\exp(-\gamma \cdot (\tau - 1))$

where $\gamma$ is rate.

Usage

risk_model_duration(duration)
risk_model_duration(duration)

Arguments

rate

The rate with which the risk dissipates

Value

A risk model

Examples

drug_history <- c(1, 0, 1, 0, 0)

risk_model <- risk_model_withdrawal(rate = 1.2) 
risk_model(drug_history) 
drug_history <- c(1, 0, 1, 0, 0)

risk_model <- risk_model_withdrawal(rate = 1.2) 
risk_model(drug_history)

Risk Model 'Long Time After'

Description

A risk model reflects how the probability of suffering the ADR changes with the drug exposures of a patient. It returns 0 when the drug prescription history has no effect, and 1 when the patient is at maximal risk. In this case, the risk of the ADR increases only after a certain moment in time before it reaches the maximal risk. Let $t_0$ be the time point that the drug was prescribed for the first time, and $\delta$ be the number of time points since first description that the function returns 0.5 ("half way"). The return value is

$1 / (1 + \exp(-\gamma \cdot (t_0 - \delta)))$

where $\gamma$ is rate. Note that this is a sigmoid function.

Usage

risk_model_long_term(rate, delay)
risk_model_long_term(rate, delay)

Arguments

`rate`	The rate with which the risk increases
`delay`	How long it takes before the risk is .5

Value

A risk model

Examples

drug_history <- c(1, 0, 1, 0, 0, 1, 0, 0, 0, 1, 1, 0, 1)

risk_model <- risk_model_long_time_after(rate = .5, delay = 9) 
risk_model(drug_history) 
drug_history <- c(1, 0, 1, 0, 0, 1, 0, 0, 0, 1, 1, 0, 1)

risk_model <- risk_model_long_time_after(rate = .5, delay = 9) 
risk_model(drug_history)

Risk Model 'No Association'

Description

Usage

risk_model_no_association()
risk_model_no_association()

Value

A risk model

Examples

drug_history <- c(1, 0, 1, 0, 0)
risk_model <- risk_model_no_effect() 
risk_model() 
# -> 0
drug_history <- c(1, 0, 1, 0, 0)
risk_model <- risk_model_no_effect() 
risk_model() 
# -> 0

Risk Model 'Withdrawal'

Description

$\exp(-\gamma \cdot (\tau - 1))$

where $\gamma$ is rate.

Usage

risk_model_past(past)
risk_model_past(past)

Arguments

rate

The rate with which the risk dissipates

Value

A risk model

Examples

drug_history <- c(1, 0, 1, 0, 0)

risk_model <- risk_model_withdrawal(rate = 1.2) 
risk_model(drug_history) 
drug_history <- c(1, 0, 1, 0, 0)

risk_model <- risk_model_withdrawal(rate = 1.2) 
risk_model(drug_history)

Risk Model 'Withdrawal'

Description

$\exp(-\gamma \cdot (\tau - 1))$

where $\gamma$ is rate.

Usage

risk_model_withdrawal(rate)
risk_model_withdrawal(rate)

Arguments

rate

The rate with which the risk dissipates

Value

A risk model

Examples

drug_history <- c(1, 0, 1, 0, 0)

risk_model <- risk_model_withdrawal(rate = 1.2) 
risk_model(drug_history) 
drug_history <- c(1, 0, 1, 0, 0)

risk_model <- risk_model_withdrawal(rate = 1.2) 
risk_model(drug_history)

Package 'expard'

Help Index

Apply Function to Subset Drug History

Description

Usage

Arguments

Value

Creating a 2×22 \times 22×2 Table

Description

Usage

Arguments

Value

Ways to construct the table

By individual time-points (time-point)

By individual patients (patient)

By individual patients (drug-era)

References

See Also

Examples

2×22 \times 22×2 Tables

Description

Usage

Arguments

Details

Value

Ways to construct the table

By individual time-points (time-point)

By individual patients (patient)

By individual patients (drug-era)

References

See Also

Examples

First Prescription

Description

Usage

Arguments

Details

Value

Examples

Compute Frequency Table

Description

Usage

Arguments

Value

expard: A package for simulating electronic health care data and fitting models

Author(s)

Fit a Model to Cohort Data

Description

Usage

Arguments

Value

Patient models

Cohort data object

See Also

Examples

Returns ADR History

Description

Usage

Arguments

Value

Generate a Cohort

Description

Usage

Arguments

Value

See Also

Generate a Drug-ADR pair

Description

Usage

Arguments

Value

See Also

Examples

Generate Drug Prescription History

Description

Usage

Arguments

Value

Examples

Generate a Patient

Creating a $2 \times 2$ Table

By individual time-points (`time-point`)

By individual patients (`patient`)

By individual patients (`drug-era`)

$2 \times 2$ Tables

By individual time-points (`time-point`)

By individual patients (`patient`)

By individual patients (`drug-era`)

`expard`: A package for simulating electronic health care data and fitting models

Print function for `generate_cohort`