Skip to content

Latest commit

 

History

History
93 lines (73 loc) · 3.68 KB

readme.md

File metadata and controls

93 lines (73 loc) · 3.68 KB

BRFSS Confounder Preparation

Ben Sabath June 22, 2021

This directory contains the pipeline to acquire information from the CDC’s BRFSS data set, which we use to acquire county level information on mean BMI and smoking habits. All data has been linked from the county level to the zip code level.

Data Provided

  • Input BRFSS will need to be downloaded independently of the code provided in this repository. Users should download the survey data from 1999-2012, following the instructions in the Inputs section.
  • All other data used is provided in this repository.

Inputs

As input to this pipeline, we use the CDC Behavioral Risk Factor Surveillance System (BRFSS) individual survey results from 1999-2012. We stop at 2012 because survey responses were no longer tagged with county level identifiers after that year.

The directory raw_data contains all of the data used to create our BRFSS confounder data, with the exception of the actual BRFSS data, which is larger than github’s max file size limit. The BRFSS data can be downloaded from here. To prepare it for the pipeline, please download the “SAS Transport Format” data for each year and place the unzipped XPT file in the directory of its respective year. The processing code will convert the XPT files first to the standard SAS format (in the sas_files directory) and then to csv format (in the csv directory).

Other files in raw_data are as follows: - state.txt: Used to convert state FIPS codes to standard postal codes - esri_zipcode_2010.csv: Used to link the county level values

Software Versions

  • R Base = 3.6.2
  • dplyr = 1.0.6
  • SAS = 9.4

Data Preparation

All code used to prepare the data is in the code directory.

To create the final data sets, first we run 1_extract_xpt.sas which converts the downloaded and unzipped .XPT files to standard SAS format files (sas7bdat). We then run 2_sas_convert_to_csv.sas to create csv files from the SAS files, to ease the process of working with the data in R.

Finally, we run 3_prepare_brfss.R which calculates the zip code level variables for use with our health data. In its original form, the BRFSS data is stored as individual survey results, which include both direct questions, and calculated variables. Information on the location of individuals is limited to their county. To create the zip code level data, we aggregate the variables of interest within each county, and then link the county level data to zip codes (using raw_data/esri_zipcode_2010.csv). We also perform moving average temporal interpolation to reduce missingness. We interpolated using a moving average within each zip code to fill in data for years where there was no survey data for a given year. We also use the same rolling average method in create an estimate for the years following 2012, when county level data became unavailable. The uninterpolated data (processed_data/brfss_confounders.csv)is kept to help control for uncertainty in the interpolation method, and allow users to identify which values were interpolated.

Output

There are two files output from this process, both of which are stored in processed_data:

  • brfss_confounders.csv: The uninterpolated zip code level BRFSS confounders
  • brfss_interpolated.csv: The interpolated zip code level BRFSS confounders, this is what we use for our analysis.

Variable details:

  • fips: 5 digit county code
  • smoke_rate: % of the respondents in a county who have ever smoked
  • mean_bmi: mean bmi of respondents in a county
  • year: Year of response
  • ZIP: zip code

Diagram of the workflow