Safety from trauma predictor update (2023 counties) and subgroup exploration

[jwalsh28]

Metric Update

This issue updates the Deaths caused by injury per 100,000 people metric. Please read through the instructions below carefully. The check at the bottom of this issue should all be completed and marked off prior to creating a final PR for the metric update.

Update goals:

1. Look into underlying cause of death CDC data (see data release schedule notes) and determine if we should shift to single race from bridge and if we can at least get overall county values
2. Update 2023 for counties
3. Consider adding race and/or age subgroups if possible
4. Consider whether we need/want to recalculate prior years data or if we handle this in another way.

Please review the instructions on the wiki before starting work on a metric update.

Crosswalks

Crosswalks were updated prior to the 2024-25 metric update process. If you are editing an existing program please check that the crosswalk being used is the latest version. For more information on crosswalks, see the crosswalk page on the Wiki.

Checklist

The checklist below outlines key steps that should be taken during the process of this metric update. These steps should all be checked off prior to finalizing the metric update.

Setup

• Metric lead has checked out a new branch from the Version2025 repo that is named after the number associated with this issue, i.e. iss###
• Metric lead has filled out the final data expectations from located in the functions folder of the repo and saved this form in the metrics data folder for all relevant final output files
• Metric lead has read through the existing version of the program and has located and examined the existing output files

Program Documentation

• The update program includes a description at the start with the date, the latest changes made and the author of the metric lead that made them
• If the program reads in raw data that is not available through an API, then the code includes a note on where this data is in Box (including the title of relevant files)
• Each step taken in the calculation is clearly documented in the code using comments
• The program is broken out into manageable steps and the code avoids using extensively long lines connected via pipes or pipe equivalents if not using R

Quality Control

• The program includes visuals of the distribution of key analysis variables throughout the calculation steps
• The program includes visualizations of the final data as well as summary statistic and a selection of assumptions tests (including count of rows by year, missing values and calculation of outliers (min & max) plus any additional checks deemed necessary)
• Assumption tests are applied to all years being created in the program. If a year in the historical version of the metric data is not being recreated in this program, that historical data is read in and differences with new years are visualized.
• The program includes the creation of a quality variable for the metric and documents the method for assigning quality grades

Reproducibility

• The program runs from start to finish without stopping due to errors or incompleteness
• The program avoids hardcoding local file paths and instead uses global paths that will work regardless of where the program is being ran (i.e. here::here() for R users)
• The program includes a “House Keeping” section which loads all necessary packages at the top of the program

Final Data

• The program reads out a final file in the form of a CSV document or multiple CSVs into a data folder in the relevant metric folder
• Final files include the relevant years in title if the metric has multiple files separated by year
• All final files being read out by the update program are put through the evaluate final data function

Review

• When ready for review the metric lead has submitted a PR to Version2025 using the PR template

[MaureenSarver]

Race categories aren’t comparable before/after 2020 in CDC WONDER, but that we are switching from age-adjusted rates to crude rates, which are much less useful for comparisons across time and geography.

[MaureenSarver]

With the bridged-race mortality data, we used 5-year pools (e.g., the 2020 measure combines data from 2016-2020). The single-race data is only available for 2018-2022, so we would only be able to do a 4-year pool for 2021 but could do a 5-year pool for 2022.

[cdsolari]

Based on discussions with @vpancini, We are not adding age subgroups, given the troubles we have had with the race and shifts to crude mortality rates (from age-adjusted rates). We will recalculate past years such that they are all based on crude mortality rates so that we can have comparability over time. Note that the data in the current dashboard with the age-adjusted data, will be preserved in an archive on the data catalog, along with archived table notes so that we can We will have to shift to the single-race data, given that the bridged-race data is no longer available. We will take the single race data back to 2018, or the earliest year available starting in 2014. We will drop the past years of data that would not be comparable to the current single race crude mortality rate data.