[BUG] cohort incidence may report incidence rate = infinity

[gowthamrao]

I observed values of 'Inf' in the output file. This should not happen. Its happening when a persons denominator start_date = end_date make denominator = 0

CohortDiagnostics/inst/sql/sql_server/ComputeIncidenceRates.sql

Line 47 in ea6b5f8

SUM(CAST(DATEDIFF(DAY, start_date, end_date) AS BIGINT)) / 365.25 AS person_years

I think we should add where clause to remove records that are contributing 0 to the denominator? Plus, if the start_date - end_date then they probably cannot contribute to rate

[azimov]

I agree that the result returned should never be inf but the datediff function should always return a number and the rate is fixed to 365.25. 0/365 = 0 - which is 0 - not undefined so this is a really result weird. The only way this ratio can result in Inf is if the DATEDIFF is Inf - which i can believe happens on some platforms if one of the dates is null maybe?

Is this on redshift or another platform?

[gowthamrao]

redshift