Continuity of Care

[Troels Kristensen]

Dear members
I 'm in the process of analysis continuity of care for patients in primary care. We have lines of data on the following 4 variable for each patient:

Personal id * GP Provider number * start date * end date

If the patient has moved from one GP to another an "end date" and related new "start date" at another clinics are generated in a new data line.
This means the data comprise from one 1 to many lines per patient.

Do you have any suggestions for statacode that has been developed to analyse this type of continuity of care?

I hope you can help me review some of the best "state of the art" in stat-coding within this specific topic? or just help me get started using useful code?

The above mentioned data are merged to a range of other data.

Thank you in advance.

All the best
Troels

[Carlo Lazzaro]

Troels:
the -group- function available from -egen- can be a good place to start from.

[Troels Kristensen]

Hi Carlo
Thank you. I will look into that. Are you aware of any existing stata code in this field of research that could be useful to explore? So far I have not been able to locate any here at statalist, but I think someone must have experience in this area :-)
Best
Troels

[Troels Kristensen]

Hi Carlo

I think egen is applied in the document on: Identifying runs of consecutive observations in panel data

Is this the type of application you are referring to? Not sure this is applying the group-function?

https://www.stata.com/support/faqs/d...-observations/
Best
Troels

[Carlo Lazzaro]

Troels:
not quite. -egen- has many functions that can be really useful.
That said, unfortunately the -egen- -group- function cannot be byable and, as such, cannot support you.
Another approach that springs to my mind is reporeted in the following toy-example:

. set obs 6
number of observations (_N) was 0, now 6

. g id=1 in 1/3

. replace id=2 in 4/6

. bysort id: g year=2020+_n

. g GP=1 if id==1

. replace GP=1 in 4


. replace GP=2 in 5/6

. egen wanted=group(id GP)

. bysort id: egen check=mean(wanted)

. list

     +------------------------------------+
     | id   year   GP   wanted      check |
     |------------------------------------|
  1. |  1   2021    1        1          1 |
  2. |  1   2022    1        1          1 |
  3. |  1   2023    1        1          1 |
  4. |  2   2021    1        2   2.666667 |
  5. |  2   2022    2        3   2.666667 |
     |------------------------------------|
  6. |  2   2023    2        3   2.666667 |
     +------------------------------------+

.

The underlying idea is that, if the id-specific mean differs from -wanted- value(s), the -id- has signed up to a different GP.