Looking for some tips and ideas.

I get a list every day of the number of appointments for each day for the next two weeks for a clinic. I have quite good history of these list, and the actual number of visits that were seen each day. So for each day I can look at how many visits were actually seen, the total number of appointments scheduled for that day one day earlier, two days earlier, ..., back 14 days earlier.

My desire is to develop forecasts for the actual number of appointments they can expect for the next 14 days.

The simplest answer I can come up with is to develop seasonal time series models based on the actual visits and use the future appointments as a floor (e.g. if 10 appointments are scheduled for a 7 days from now, I should not predict that they will only have 8 appointments).

But, this is not ideal, as it does not take into account that if there are 10 appointments 1 or more may not show. I would somehow like to incorporate the history data I have into the forecasts. Using the future appointments as factors does not seem like a good ideas as they are not independent.

I have not found any examples of anyone performing forecasts with this type of data. Is there a way to potentially use this data in a better way?

FYI: I am not necessarily bound to any technology, I can use either R or SAS.

  • $\begingroup$ Do you have some general plot of the data? People miss appointments for several reasons: weather, day of week, age and maybe even gender. If your original data can explain the target values well, just try some kind of regression. $\endgroup$ – Fernando Mar 13 '14 at 18:33

Can there be more actuals than scheduled?

This sounds like a simple regression problem. Y(actual) and X(scheduled).

Can you post your example data with your history of scheduled and actuals?


Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.