Changing patients IDs from string to numeric

[Fouziah Almouqati]

Hi all,

I am working on a big dataset that includes more than a million patients with unique fake IDs. The IDs take the form of "map_code_111000" and are stored as strings. I want to convert them to a numeric format. What would be the best approach to tackle this? I was thinking of either removing the "map_code_" part or extracting the number "111000". Is this an acceptable approach? If yes, what function or syntax should I use? If not, what approach do you recommend?

[Joseph Coveney]

Among the safest ways:

assert !missing(patient_id)

preserve

contract patient_id, freq(pid)
quietly replace pid = _n // Stata will promote to -long- storage type
assert !missng(pid)

tempfile patient_list
quietly save `patient_list'

restore
merge m:1 patient_id using `patient_list', assert(match) nogenerate noreport

If you want to keep your patient ID crosswalk table for later reference, then change the temporary file in the code snippet above to a permanent one.

[Joseph Coveney]

You can also use egen . . . tag() after sorting the dataset, and follow-up with the sum() function. You won't automatically get a crosswalk table as an intermediate result. If that's important to you. then you can get it afterward with a bysort patient_id: keep if _n == 1.