Sentinel Routine Querying Tool Report Documentation
Public
Source
5
5
Regardless of the study type performed, there will inevitably be loss of study members over time. While the QRP can perform statistical techniques to account for this loss-to-follow-up, users may desire to visualize the impact of censoring. Depending on the study design, the QRP reporting tool provides two types of figures to fit this purpose. In the context of these visualizations, "censoring" refers to any reason that leads to the end of observed time.6
6
7
7
For studies where exposed and unexposed members are followed over time for the occurrence of a user-specified outcome (i.e., Exposures and Follow-Up Time [Type 2] and Medical Product Switching Patterns [Type 6]), the reporting tool outputs Kaplan-Meier (K-M) curves. The QRP reporting tool plots follow-up time in days on the x-axis and the cumulative conditional probability that the outcome of interest has not occurred on the y-axis. This probability is calculated using the K-M estimator of the complement of the cumulative distribution function (i.e., 1-CDF) as described in the "Calculation" section below. Because the cumulative probability is conditioned on all reasons for censoring and presumably not all study members experience the outcome of interest, the K-M estimator of the complementary cumulative probability will start at 1 when the study begins and end somewhere above 0 at the end of the study.8
8
9
9
For studies where no outcome is specified formally, users can still visualize the cumulative probability that a given censoring reason has not occurred over time. In contrast to the K-M curves described above, these figures plot unconditional probability; in other words, the probability that a study member has not experienced the censoring reason of interest at each point in time among members who do experience that censoring reason. Because of this, these figures (described as "1-CDF" in the table below) display a cumulative probability of 1 at time 0 and cumulative probability of 0 at the end of the study.10
10
11
11
Using parameters in `qrp_report.FigureFile`, users can specify which figures they would like output as well as which censoring reasons to display and the way in which to display them on each figure.12
12
13
13
The user can specify in the `qrp_report.GroupsFile` which cohort groups should be displayed for Type 1, descriptive Type 2, and Type 5 analyses, as well as which analysis groups should be included for Type 6 switching analyses.14
14
15
-
For inferential Type 2 analyses using unweighted propensity scores (PS), the user can output unadjusted (both conditional and unconditional) K-M figures, as relevant for the chosen analysis type. The user does not have an option of which analysis groups to include.15
+
For inferential Type 2 analyses using unweighted propensity scores (PS), the user can output unadjusted K-M figures, as relevant for the chosen analysis type. The user does not have an option of which analysis groups to include.16
16
17
17
|**Inferential Analysis Type**|**K-M figures available**|18
18
|-----------------------------|-------------------------|19
19
|1:n fixed-ratio PS-matched analysis|<ul> <li>Unadjusted</li> <li>Adjusted conditional</li> <li>Adjusted unconditional</li> </ul>|20
20
|1:n variable-ratio PS-matched analysis|<ul> <li>Unadjusted</li> <li>Adjusted unconditional</li> <ul>|21
21
|Unweighted PS stratified analysis|<ul> <li>Unadjusted</li> <ul>|22
22
|PS stratum weighted analysis|<ul> <li>None</li> <ul>|23
23
|Inverse probability of treatment weighted (IPTW) analysis|<ul> <li>None</li> <ul>|24
24
|Covariate stratified analysis|<ul> <li>None</li> <ul>|25
25