Reports | Digit Preference

Digit Preference
This analysis assesses the use of terminal digits (either first or last in numerical findings) by study sites when reporting their clinical findings. It can be used to identify those sites that might exhibit biases in rounding issues or other problems with how they report data as compared with all other sites in the study.
This analysis assesses the use of terminal digits (either first or last in numerical findings) by study sites when reporting their clinical findings. It can be used to identify those sites that might exhibit biases in rounding issues or other problems with how they report data as compared with all other sites in the study.
Running Digit Preference for Nicardipine using default settings generates the report shown below.
Digit Preference Volcano Plot
In this Volcano Plot , e ach point represents the comparison of a site to all other sites. This comparison is used to determine whether there is a difference in distribution for the last digit for a findings test with numeric data available and is done for all sites across all tests in all findings domains.
The Y axis is the -log 10 (Raw Row Mean Score p-value ), which takes advantage of the ordinality of the final digit value, This test uses standardized midrank scores in case there are gaps due to certain digit values not present. Midranks are a way of scoring the columns when the distance between levels does not necessarily have a practical interpretation. Large numbers on the Y axis indicate statistically significant results.
The X axis is the maximum percent difference 1 across all digits between a site versus all sites.
Values far from 0 indicate important differences between a site and the reference distribution of all other sites. An FDR ( alpha =0.05) line is indicated by the dotted red line. Values above this line can be considered significant adjusting for multiple comparisons. This could identify rounding issues or other problems with how a site reports a particular test compared to other sites.
Test Results
The Laboratory Test Results section contains the following elements:
Shows the number of tests performed at each study site, the numbers of each type of test, and the subjects delineated by the specified digit (first or last) of the subjects’ USUBJID. Selecting any group (for example, all tests for subjects whose USUBJID ends with “9”) as shown above, highlights those tests across all distributions. There is a separate set of distributions for each Findings domain.
One Data Filter .
Enables you to subset subjects based on study site, test, and digit. Refer to Data Filter for more information.
Action Buttons
Action buttons, provide you with an easy way to drill down into your data. The following action buttons are generated by this report:
Show Sites : Shows the rows of the data table for the selected points from the volcano plot . Use your mouse to select one or more sites of interest before clicking this button, as shown below:
Clicking opens the following table:
Digit Bar Charts : Clicking displays a bar chart , comparing the last digit distribution between selected sites versus all others, for the points selected in the table. This gives the user the ability to compare just how different each site is for a particular test. The underlying data table is available by going to Script > Data Table . The following chart shows the sites/tests selected above:
In this example, site 16 shows a marked preference for using reporting diastolic pressure with a terminal digit of “0”.
General
Click to generate a standardized pdf - or rtf -formatted report containing the plots and charts of selected sections.
Click the Options arrow to reopen the completed report dialog used to generate this output.
Click the gray border to the left of the Options tab to open a dynamic report navigator that lists all of the reports in the review. Refer to Report Navigator for more information.
Methodology
Compare the observed distribution of the last or first digit for each test with each site (the suspect site, indexed with s ) compared to all other sites taken together as a reference (indexed as o ).
Compared using a row mean score chi -square tests (Stokes et al ., 2012) 2 to take advantage of the ordinality of the column variable.
Scores are based on standardized midranks , often used when column values cannot necessarily be considered equally spaced (which tends to happen if not all digits are presented).
FDR p -values are calculated and the reference line is determined as described in How does JMP Clinical calculate the False Discovery Rate (FDR)? .
Report Options
Report Option Descriptions
Specific documentation for each of the options can be viewed by clicking on the following links:
General Options
Analyze all tests from all findings domains , Findings Tests , Analyze: , Consider BY variables in the analysis
Summarize sites with at least this many subjects: , Alpha
Subject Filter 3
Additional Filters
Additional Filter to Include Subjects 4 , Select the population to include in the analysis

1
This is the maximum of ( p(0 in suspect) - p(0 in reference ), p(1 in suspect) - p(1 in reference) , … p(9 in suspect) - p(9 in reference) ) where p(x) is the percent of records, suspect is the site in question and reference is all other sites that are not the site in question.

2
Stokes ME, Davis CS, Koch GG. (2012). Categorical Data Analysis Using SAS, Third Edition. Cary, NC: SAS Institute, Inc.

3
Subject-specific filters must be created using the Create Subject Filter report prior to your analysis.

4
For more information about how to specify a filter using this option, see The SAS WHERE Expression .