Code snippets, utilities, and notes related to SAMHSA SUD dataset and project.
Only 2015-2019
The Treatment Episode Data Set (TEDS) system comprises demographic and drug history information about individuals admitted to and discharged from treatment. The TEDS system is admission-based, therefore, TEDS admissions and discharges do not represent individuals. Each row represents an admission/discharge. TEDS-A records the admissions while TEDS-D records the discharges. The two data sets are separate but linkable.
The raw data is open access and is available for download from the SAMHSA website. This repo contains code used to filter and process raw data files.
Currently only TEDS-D data files from 2015-2021 are processed.
Data Folder: /secure_net/datasets/opendata/samhsa_sud/
Data dictionary: The current data dictionary is relevant to the variables included in the 2015-2021 TEDS-D processed file on the server, and additional details can be obtained from the 2021 TEDS-D Codebook available on the SAMHSA website. Variables that were dropped from the raw data are not included in the data dictionary.
- When loading the processed csv file using pandas, must set keep_default_na=False to prevent conversion of None str into NA (see tedsd_sample_notebook)
- All true missing are coded as such, therefore there should be no na values on import.
- PREG: Pregnant at admission
- For males and females not of child-bearing age this variable should be recoded to something like "Not Applicable"
- LOS: Length of stay
- recorded as individual numbers until 31, then binned.
- Consider binning the individual numbers as well so it's all categorical
- 18 Nov 2024
- Fixed an issue where None values in 2021 were being converted to na
- Fixed IDU and MARSTAT mapping issues.
- Reprocessed the 2015-2021 raw files and generated new clean derived dataset.