You can use @Chris_Knoll 's logic above, but generally it is best to toss these records or create a record in the Observation table for ‘past medical history’ or ‘past surgical history’ for the Condition or Procedure. You know the Person had it in the past, you just don’t know the day, so it might be useful criteria for a study. You could use the date of the Visit or the date of the data extract for the Observation start date.
Measurements are tricky because the Measurement doesn’t always occur on the same day as the Visit. Many times Persons have the Measurement done a week before the Visit, so the Provider can review results and assign a course of treatment. Other times the Measurement is done on the day of the Visit. And sometimes the Measurement is taken days after the Visit because the Person needs to do certain things (like not eat for 12hrs) before a Measurement can be taken. I say toss the dateless Measurement records. The timeline and order for clinical events are very important.
If an end date can be NULL (i.e. condition_end_date) and your data does not contain an end date, then leave it NULL.
Now for drug end_dates, see the conversation here. If the drug record is an “administered drug” record, then the end date = start date. There may possibly be very few cases where this is not true (example: iv infusion started before midnight and ending after midnight), but it’s such a small percentage that it won’t affect analysis. If the assumption concerns you, then toss out all iv infusion records without an end date. For other drug end dates you could derive them by start date + days supply = end date or other calculations using the start date (has to be populated) and the quantity, dose, sig, etc.
Condition and Device end_dates are not required, so leave them NULL if you don’t have an end date. Measurement and Procedure do not have end dates.