First Review of Code Scripts

[laurenflynn]

File 1 QID Year: Line 32: why is f[2:18] selected? Line 49: the last_year_cont function logic is a little unclear. Lines 74-78: What is [, zip := ZIP][, ZIP := NULL][zip >= 501] doing?

File 2 First Hospitalization: Line 32: what is the difference between ADATE and year?

File 3 Denominator: Line 4 and 55: Why does this file filter to years 2009-2016 but files #1-2 started at 2000? Header says input of denominator data for 2000-2016.

File 5 Single state analysis: Line 40: [, first_hosp := TRUE] unclear what this is doing Line 47: why is exit restricted to after 2009? Lines 91-97: exp_data function is not clear Line 91: why is the list generated in reverse order ie 2016:2009?

File 7 Cont Us: Line 77: Should the variable name be pre2010qid? Why is the analysis restricted to exit after 2009?

[daniellebraun]

tagging @danielmork as well

[danielmork]

Thank you for the questions. See my answers below. Please let me know if there are further questions!

File 1 QID Year: Line 32: why is f[2:18] selected?

• f[1] is 1999 denominator data, did not need this file

Line 49: the last_year_cont function logic is a little unclear.

• Function to return last year during first stretch of continuous Medicare FFS enrollment.
• Here is the long version:
• function(year) # input year is sorted integer vector
• for (y in first(year):2017) {
• if (!(y %in% year)) return(y - 1)
• }

Lines 74-78: What is [, zip := ZIP][, ZIP := NULL][zip >= 501] doing?

• this could probably have been done more intuitively;
• 1) rename ZIP to zip (by creating new variable)
• 2) remove ZIP
• 3) restrict entries to zip >= 501 # valid zip codes are >= 501

File 2 First Hospitalization: Line 32: what is the difference between ADATE and year?

• ADATE is exact date of hospitalization, year is what it says

File 3 Denominator: Line 4 and 55: Why does this file filter to years 2009-2016 but files https://github.com/NSAPH-Projects/AirPollution_DistributedLag_ADRD/issues/1-2 started at 2000? Header says input of denominator data for 2000-2016.

• each individual who makes it to the study period beginning in 2010 is followed until 2016
• we need denominator data from 2000-2009 to get the zip codes for annual exposure
• we use denominator data from 2010-2016 for time-varying zip code level confounders
• originally I was going to begin the study period in 2009, hence this file beginning then.

File 5 Single state analysis: Line 40: [, first_hosp := TRUE] unclear what this is doing

• creates a variable for year of first ADRD hospitalization
• when merged with denominator data, will set this to be FALSE for everyone else

Line 47: why is exit restricted to after 2009?

• the study period begins in year 2010, if individual exits FFS prior to 2010 we will not have hospitalization data

Lines 91-97: exp_data function is not clear

• this is a confusing one and I wrote code for speed, but lost some interpretability
• 1) each 'qid_yr_exp' file has yearly exposures (columns) for each individual (rows)
• 2) for each study year (2010-2016) I want to grab the exposure history for each individual from 2000 up until that year
• 3) this results in data.tables with different numbers of columns. merging fills missing columns with NA

Line 91: why is the list generated in reverse order ie 2016:2009?

• not sure if this was necessary, since the data.tables have different numbers of columns (but the same columns, just more: see answer to previous question), then the first data.table has all the columns

File 7 Cont Us: Line 77: Should the variable name be pre2010qid? Why is the analysis restricted to exit after 2009?

• we want to remove anyone who had an ADRD hospitalization prior to 2010, this variable captures those individuals and removes them from the analysis