Possibly a very basic question, but I would really appreciate clarification from those with a strong background in stats/epi.

I have a number of human patient medical records. They are retrospective and longitudinal. Observation from 2000 to 2016 of 18 to 65 yr olds. However, some patients will not have a complete 16-year record as:

1) Their record will appear only when they turn 18, so I may only have records e.g., from 2008 for some patients.

2) A patient may die (but there won't be an indication) before 2016.

Each patient will have a number of events e.g., prescription of a drug. I would like to calculate the mean number of these events a patient will experience. However, with the patient records, not all being the same in duration I am unsure how I adjust the mean calculation.

As all ways, suggestions are very welcome.

  • 1
    $\begingroup$ The simplest method might be to use the average number of events per year based on the number of years available. This would work best if important events are, roughly speaking, uniformly distributed across years. // If there is a 'key' time period during a patient's history (e.g., within 2 years after diagnosis) when events are especially important, then perhaps use number of events in key period (if included in record) for each patient. $\endgroup$ – BruceET May 25 at 19:39
  • $\begingroup$ @BruceET I think your comment would be an acceptable answer, consider turning it into one. $\endgroup$ – Martin Modrák May 29 at 12:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.