defineMONITORvars

Input <code>data.table</code> or <code>data.frame</code>

data

The name of the unique subject identifier (character, numeric or factor).

The name of the time/period variable in <code>data</code>.

The name of the binary indicator of missingness or imputation for I at time point t. This is used for coding MONITOR(t-1):=1-imp.I(t).
When imp.I(t)=1 it means that the patient was not observed (no office visit) at time-point t and hence no biomarker was measured.

imp.I

The name of the new column that represents for each row t the indicator of being MONITORed (having a visit) at time points t+1.
This variable is added as a new column to the output dataset.
The column MONITOR(t) is set to 1 when the indicator imp.I (imputed biomarker) is 0 at time-point t+1 and vice versa.

MONITOR.name

The name of the future column (created by this routine) that counts number of periods since last monitoring event at t-1. More precisely,
it is a function of the past <code>N(t-1)</code>, where 0 means that N(t-1)=1; 1 means that N(t-2)=1 and N(t-1)=0; etc.

tsinceNis1

The name of the numeric biomarker value which determines the dynamic treatment rule at each time point t.


Helper routine to define the monitoring indicator and time since last visit


Analysis of longitudinal time-to-event or time-to-failure data.
Estimates the counterfactual discrete survival curve under static, dynamic and
stochastic interventions on treatment (exposure) and monitoring events over time.
Estimators (IPW, MSM-IPW, GCOMP, longitudinal TMLE) adjust for measured time-varying
confounding and informative right-censoring. Model fitting can be performed either
with GLM or H2O-3 machine learning libraries.
The exposure, monitoring and censoring variables can be coded as either binary,
categorical or continuous. Each can be multivariate (e.g., can use more than one
column of dummy indicators for different censoring events).
The input data needs to be in long format.

Oleg Sofrygin

stremr

Streamlined Estimation of Survival for Static, Dynamic and
Stochastic Treatment and Monitoring Regimes

defineMONITORvars function

 Convert the input long format data with the time ordering: (I(t), imp.I(t), C(t), A(t))
into the data format required by the stremr input functions: (I(t), C(t), A(t), N(t):=1-imp.I(t+1)).
N(t) at time-point t is defined as the indicator of being observed (having an office visit) at time point t+1 (next timepoint after t)
The very first value of I(t) (at the first time-cycle) is ALWAYS ASSUMED observed/measured (hence I.imp=0 for each first subject-time observation).

Details

 The input data.frame data needs to be in long format. The format of the specified columns needs to be as follows. The time ordering of the input data at each t is as follows: (I, imp.I, CENS, TRT) The time ordering of the output data at each t is as follows: (I, CENS, TRT, MONITOR), where MONITOR(t)=1-imp.I(t+1). In output data.table, MONITOR(t-1)=1 indicates that the biomarker I(t) at t is observed and vice versa. In addition the output data.table will contain a column 'tsinceNis1', where:
<ul>
<li> tsinceNis1(t) = 0 means that the person was monitored at time-point t-1.
  </li>
<li> tsinceNis1(t) > 0 is the count of the number of cycles since last monitoring event.
</li>
</ul> 

defineMONITORvars: Helper routine to define the monitoring indicator and time since last visit

Description

Usage

Arguments

Value

Details

Data Format