Authors: Amitabh Chandra, Evan Flack, Ziad Obermeyer
Contact: Evan Flack (flack@stanford.edu)
-
read_in: Scripts to read-in the raw Medicare .sas files, subsets/formats variables -
00_pre_process: Scripts to start other scripts, define functions/plot themes -
01_sample: Scripts to create the main analytic sample, along with prediction and falsification samples -
02_features: Scripts to create features used in the analysis, such as mortality outcomes and initial-90 days spending -
03_model01_prediction: Scripts used to predict risk outcomes in the prediction sample of dual eligible beneficiaries 66+02_iv: Scripts to make tables/figures for the IV analysis in the main text and appendix
-
sup_data: Data used in addition to the Medicare claims stored on the NBER aging servers
This project used Medicare claims/administrative data made available though a data use agreement (DUA) with the Center for Medicare and Medicaid Services (CMS). It is stored on the NBER aging servers. The raw data is stored in SAS files; the aging servers are set up such that only a few are able to run SAS programs. Because of this, the analysis cannot be run with one shell script. Instead, there are a series of shell scripts, alternating between .SAS, and .R programs.
First, you will need to set the directories where you would like data to be stored. This is done in sas_librefs.sas and 00_pre_process/start_script.R. The "lib_base_data" object from start_script.R must be the same as both "em" and "exp_dir" in sas_librefs.sas. The "lib_base" object in 00_pre_process/start_script.R should be the root directory of the scripts. Also move the data stored in sup_data to the "lib_base_data" directory.
read_in/00_run_sas_scripts.sh01_sample/00a_run_sas_scripts.sh01_sample/00b_run_r_scripts.sh01_sample/00c_run_sas_scripts.sh02_features/get_drug_info/00_get_drug_info.sh02_features/00_run_r_scripts.sh03_model/01_prediction/00_predict_risk.sh03_model/02_iv/00_run_iv_scripts.sh