Abstract
Background/Aims GHRI has created a SAS program that takes all of its input from a configuration file. The configuration file is nothing more than a text file, formatted to easily convey the datasets, variables, and criteria necessary for data extraction.
Methods There are numerous benefits of this approach: (1) complex logic can be encapsulated in a single file; (2) extensive comments in the configuration file are allowed; (3) the program is “safe” for multi-site requests; (4) after review, the program does not need to be redistributed, only the configuration file needs to be distributed and reviewed for each request, and it cannot execute without the program that supports it; (5) summary-level datasets with sophisticated criteria can be created; and (6) current summary-level efforts prevent person-level analysis.
Results Here is how the program works: The configuration file consists of different sections. The first section tells the program which VDW datasets are to be used. Other sections create criteria for selection and analysis. For example, here are two criteria for diabetes identification.
# outpatient criteria criteria_name=outpatient / criteria_where=enctype not in (“IP”,”ER”) and (dx in(‘357.2’,’366.41’,’362.01’,’362.02’,’362.03’,’362.04’,’362.05’,’362.06’, ‘362.07’,’648.83’) or substr(dx,1,3) eq ‘250’) / criteria_keys=dx dx_date / criteria_data=dx dx_date / criteria_sort=a / # lab a1c criteria criteria_name=lab_a1c / criteria_where=test_type eq “HGBA1C” and input(result_c, 8.) gt 6.5 / criteria_keys=test_type result_c lab_date / criteria_data=test_type result_c lab_date / criteria_sort=a /Conclusions Multiple criteria blocks can be assembled to create flags for each person in a cohort. The resulting summary-level dataset can then be used to make determinations based on the absence or presence of particular flags. The criteria blocks use simple boolean expressions and concentrate the “intelligence” of the analysis in one location.
