readmission2: Transformed Readmission Data for Illness-Death Modeling

In frailtypack: Shared, Joint (Generalized) Frailty Models; Surrogate Endpoints

Transformed Readmission Data for Illness-Death Modeling

Description

A dataset derived from the readmission data (originally from the frailtypack package, related to rehospitalization times after surgery in colorectal cancer patients). This transformed version reshapes the data to fit a standard illness-death model framework, focusing on the first event (rehospitalization) and the terminal event (death). Recurrent rehospitalization events beyond the first one are excluded. Time is scaled to years.

Usage

data(readmission2)

Format

A data frame with one row per subject, containing columns suitable for use with the IllnessDeath function:

id

Unique subject identification number.

observed_disease_time

Time (in years since surgery) to either the first rehospitalization (illness), death, or administrative censoring, whichever occurred first.

observed_death_time

Time (in years since surgery) to either death or administrative censoring.

disease_status

Indicator for the non-terminal event (first rehospitalization). 1 if the subject experienced a first rehospitalization before death/censoring, 0 otherwise.

death_status

Indicator for the terminal event (death). 1 if the subject died, 0 if censored.

dukes

Dukes' tumoral stage at baseline (Factor or numeric: 1:A-B; 2:C; 3:D).

sex

Gender (Factor or numeric: 1:Male; 2:Female).

charlson

Comorbidity Charlson's index at baseline (Factor or numeric: 0: Index 0; 1: Index 1-2; 3: Index >=3). Note: Original data had this as time-dependent, this version likely uses the baseline value.

chemo

Indicator whether patient received chemotherapy (Factor or numeric: 1:No; 2:Yes).

group

An example grouping variable (numeric, derived from id mod 10 + 1), useful for fitting grouped frailty models.

Details

The transformation process involved:

1. Starting with the original readmission data.

2. Excluding recurrent rehospitalization events, keeping only the interval from surgery (t.start=0) to the first event (event=1) or censoring (event=0).

3. Reshaping the data so each row represents one subject.

4. Defining observed_disease_time and disease_status based on the first event interval (t.stop when t.start=0).

5. Defining observed_death_time and death_status based on the overall follow-up time and final death status for the subject. If a subject had a first event and then further follow-up, the death time comes from the second interval if available.

6. Scaling time variables (t.stop) from days (assumed) to years by dividing by 365.

7. Copying baseline covariates (dukes, sex, charlson, chemo) from the subject's first record.

This dataset is intended primarily for demonstrating the IllnessDeath function.

Source

Derived from the readmission dataset, originally described in: Gonzalez, JR., Fernandez, E., Moreno, V., Ribes, J., Peris, M., Navarro, M., Cambray, M. and Borras, JM (2005). Sex differences in hospital readmission among colorectal cancer patients. Journal of Epidemiology and Community Health, 59, 6, 506-511.

frailtypack documentation built on Nov. 23, 2025, 1:09 a.m.