README.md

In stephematician/trialcostR: Trial Cost-Effectiveness in R

trialcostR

Stephen Wade

04 April 2017

Introduction

An R package to assist in calculating cost-effectiveness analysis of a trial.

Provides three classes; - Trial, - ExtendTrial, and - UtilityTrial, and provides generics for calculating (with or without discounting); - life expectancy (lifeyear() and dlifeyear()), and - quality-adjusted life expectancy (dqaly()).

The base class Trial does not allow for any beyond-trial extension or censoring, and so life expectancy is given by the follow-up time, or discounted from the follow-up time.

The ExtendTrial class allows for a new follow-up time to be specified and accepts hazard data so that the life expectancy up to the new follow-up time can be estimated. Both censored and uncensored observations are accepted, where uncensored observations' (discounted) life-expectancy are given by the original (discounted) follow-up times.

The UtilityTrial class further allows for the specification of health states and the utility of those health states.

Discounting is applied continuously by exponential functions.

Hazard data is allowed to vary by outcome, and the hazard is assumed constant on each time interval specified, but is otherwise treated as a continuous function of time. Time interval data is continuous and not restricted to any discrete time scale (e.g. whole years etc).

Health state utilities, like hazard data, can vary by outcome, and are assumed constant within each health state time interval. The time interval data is continuous as per hazard time intervals.

All calculations are performed by exact integrals of, essentially, step-like functions for hazards and health state utility.

Usage

Given long-data from a trial, a Trial class is created by specifying the columns that describe participants, the columns that describe the outcomes and the column that contains the follow-up time

A trivial example is

    df <- data.frame(age=sample(c('55-59', '60-64'), 10, replace=T),
                     gender=sample(c('male', 'female'), 10, replace=T),
                     alive=sample(c(T, F), 10, replace=T),
                     fup_time=10*runif(10))
    discount_rate <- 0.03

    my_trial <- Trial(df,
                      ~age + gender,
                      ~alive,
                      fup_time,
                      discount_rate)

Accessor functions for the participant data, outcome data and follow-up data are given

    participants(my_trial)
    outcomes(my_trial)
    lifeyear(my_trial)
    dlifeyear(my_trial)

The name lifeyear implies that the follow-up time was measured in years, however the Trial class does not interpret the follow-up data in any form of units. The lifeyear function is trivially returning the follow-up times.

The discounted life-year function dlifeyear returns the (continuously) discounted life-years of follow-up.

Beyond-trial life expectancy

The ExtendTrial class provides functions to estimate the beyond-trial life expectancy, either discounted or undiscounted.

Beyond-trial quality-adjusted life expectancy

This code is licensed under the MIT license and you may use this package strictly under those terms.

stephematician/trialcostR documentation built on May 30, 2019, 3:18 p.m.