Zero Mortality Incidence


I have a question regarding All Mortality Incidence estimates. I am working in the US setup (BenMAP Version 1.5) and have a user-defined grid of census block groups within a 4-county area. My model inputs for baseline PM2.5, control PM2.5, and population data all have 2,647 row-column combinations (which represent 2,647 census block groups in my grid). The population data is an output from the PopGrid program and I have included the _popwgt file in my population dataset to show what percentage of each census block group falls within each county. I have double checked that the row-column combinations for these inputs match the row-column attributes from the shapefile I used to define my grid domain.

As shown below, my air quality delta also populates all 2,467 census block groups:

However, when I run the EPA Standard health impact functions for PM2.5 (based on the BenMAP 2030 mortality incidence dataset), incidence data is shown for all but about 200 of the census block groups defined in my grid domain (this is true for all the standard PM2.5 HIFs, regardless of age groups represented). A small portion of these domains that are missing incidence data are explained by having zero population or no delta, but about 160 have both population (for the age groups specified in the HIFs) and nonzero baseline-control model data.

What are some possible reasons that I would get no incidence data (point estimates) in my health impact results for certain row-column combinations in my user-defined grid? These row-column combinations aren’t even shown as having zero values in the health impact results data table, they are just omitted altogether.

Thank you!

Hi Kate,

I’m sorry to hear that you’re having trouble generating incidence estimates. Before suggesting a few ways to address this issue, let me first thank you for including such clear screenshots and a very detailed description!

So, here’s where I would start:

  1. Confirm that your cfgrx results file is reporting population and a baseline incidence value to each census tract. You can check this by double-clicking one of your results under the “Health Impact Results” tab and then select “Data” to see your results as a table. The columns marked “population” and “baseline” should each be populated with values.
  2. You might also want to regenerate the cross-walk file. The program assigns data from one spatial scale to another in its GIS; it creates a “percentage cross-walk” file that tells it how to weight the values in one grid cell when assigning those values to another another. Sometimes this file gets corrupted. However, you can rebuild the file by navigating to tools–>compute grid crosswalks.

I hope this helps. Let me know what you find!

Hi Neal,

Thank you so much for getting back to me!

  1. Confirmed that the cfgrx file is reporting population and baseline incidence for a portion of the census block groups in my analysis (this file still excludes the ~200 groups as discussed in my original message).
  2. I regenerated the crosswalk file but this did not fix the issue.

I used the database export tool to export my population data and realized that the ~200 grid domains missing were not included in this file. However, the population data that I loaded into the Populations Datasets includes these ~200 missing domains.

I assume that this means that there is something wrong with the connection between my user-defined grid and population data for these ~200 missing domains… Do you have any suggestions of how to resolve this? At least in tabular form, my user-defined grid row-column identifiers match all of the row-column identifiers in the population dataset that I loaded into BenMAP.

Thanks again,

Hi Kate,

Thanks for sharing these additional details. Given that you’ve compared the census block group population file index and the census block group shapefile index, and these appear to match, I wonder if there is something unusual about the indices for these ~200 or so rows. Can you copy an image of these non-matching rows and paste them into your response?


Hi Neal,

I’m not sure what you specifically want me to show, but here is a screenshot of the PopGrid output population data for a subset of the ~200 mismatched rows (this is what I input as my population data):

Here is the a subset of the _popwgt file from the PopGrid output for the ~200 missing rows:

Here is the attribute table for my user-defined grid and all of the row-col listed are part of the ~200 mismatched rows:

I am using Woods and Poole growth rates to grow the population data from 2010 to 2030. Could there be an error with the growth rates?

Here is a screenshot of what I have been using for population input:
Validation passed with 0 warnings.

Please let me know if there is anything else I can provide.


Hi Kate,

Thanks, this is helpful. I’m wondering perhaps if the program won’t accept a Column value that is that long. Are the other (non-omitted) tract FIPS values as long as the ones pasted above?

Hi Neal,

Thanks for getting back to me. Yes - all of my column values are 9 digits long, not just the ~200 missing from the population data export.

I resolved this issue by manually changing my row, column, and population data (including the _popwgt file) from “general” to “number” in excel. Prior to this, I had been directly inputting the PopGrid output into BenMAP, assuming that it was in the correct format. Thanks for all of your help!