Patient Cohorts - KRT

Aim

To produce an incident cohort in which the only limiting factors are the accuracy of the data.

Algorithm - Incident/Prevalent

Step I - Produce Initial cohort

For a RDA sending facility select candidate patients on the following criteria:

Patientrecord sent by the faculty rda feed.

UKRDC treatment records overlapping a period 90 days prior to the beginning of the time window and 90 days post the end of the window.

Patient must be alive or have a date of death greater than the beginning of the window.

Step 2 - Chain treatments together

For each patient treatment append the the start of the next treatment

This is incomplete, and requires edge cases properly thought about. What exactly do we mean by next treatment. How do we properly think about treatments that overlap

Step 3 - Remove treatments based on recovery windows

Recovery events are identified by treatment gaps greater than 90 days.

If recovery occurs on or after the end of the window drop all following treatments. This is a patient who is in recovery in the following quarter.

If recovery overlaps with the end of the window remove patient entirely. Patient recovered during the quarter and so shouldn’t be included as incident.

Step 4 - Build treatment timelines

Recode acute patients to chronic. If a patient ever becomes chronic recode them as chronic in all prior treatments.

Calculate the start of the treatment timeline from the start time of the first treatment of each patient. Note that at this point all recoveries for greater than 90 days have been removed so any short treatments and recoveries can be summed.

Calculate the end of the treatment timeline from the to time of the last treatment. The timeline length is defined as the time between the timeline start and the timeline stop. The life length is also calculated from the time between the timeline start and the time of death.

Step 5 - Assign incident/prevalent

Prevalent patients are those who are coded as ckd, have a timeline length greater than 90 days, and whose timeline overlaps with the end of the window (the prevalence point).

Incident patients are those which have a timeline start within the window and meet one of the following criteria:

Remain on KRT for greater than 90 days

Die within 90 days but are on a chronic code