ASA Connect

Do you know of any literature that provides teh method for calculating # of hospitalization per patient per month and it confidence interval.​

------------------------------
David Ipe
Pharmacyclics
------------------------------

I'm not sure what you mean re: the method for calculating # of hospitalization per patient per month - typically you just sum up the number of hospital days per person and also sum up the number of days at risk. The CI and test would depend on the distribution - so it could be Poisson, but is likely to be skewed, so something like a negative binomial or zero-inflated Poisson might fit the data better. Hope that helps.

------------------------------
Douglas Landsittel
Professor of Biomedical Informatics
University of Pittsburgh
------------------------------

Original Message:
Sent: 04-12-2019 13:44
From: David Ipe
Subject: Number of hospitalization per patient per month

Do you know of any literature that provides teh method for calculating # of hospitalization per patient per month and it confidence interval.​

------------------------------
David Ipe
Pharmacyclics
------------------------------

Hi David,
It sounds like you're working on an interesting project!

I'd quickly second Prof. Landsittel's reply that
(i) It would help to have greater detail about the exact statistic you are wanting to calculate (and perhaps greater context for the more general problem you are looking to solve).
(ii) Your question (with regard to CIs) is heavily dependent on the answer to the first.

The calculation of "number of hospitalizations" is very definition-driven. So in this regard there is no single method to do this that is "THE method". Furthermore, it would help to know if you're looking to (i) estimate this number from data you have, or (ii) simply to look-up how other people make these calculations.

Sources of variation for the definition of hospitalization rates include
(i) variation in ward
(ii) whether this was an admission vs a readdmission
(iii) whether and how readmissions are merged with the original admission.
- by proximity in time
- by underlying cause of admission
(iv) grouping related admissions to different hospitals or hospital systems

Going through these items in detail will help us get you to the right sources of information.






------------------------------
Glen Wright Colopy
DPhil Student
University of Oxford
------------------------------

Original Message:
Sent: 04-12-2019 13:44
From: David Ipe
Subject: Number of hospitalization per patient per month

Do you know of any literature that provides teh method for calculating # of hospitalization per patient per month and it confidence interval.​

------------------------------
David Ipe
Pharmacyclics
------------------------------

This is an prospective observational study and we are observing patients who receive treatment at enrollment into the study all the way till they receive the next treatment after the index treatment (i.e treatment at enrollment).
A patient may have one or more hospitalizations during this observation period and we are interested in calculating the number of hospitalizations per patient per month for the entire cohort of patients who were enrolled in this prospective observation study. Some of the patients may have zero number of hospitalizations.
Hope this clarifies.
Thanks
David

------------------------------
David Ipe
Pharmacyclics
------------------------------

Original Message:
Sent: 04-15-2019 11:32
From: Glen Colopy
Subject: Number of hospitalization per patient per month

Hi David,
It sounds like you're working on an interesting project!

I'd quickly second Prof. Landsittel's reply that
(i) It would help to have greater detail about the exact statistic you are wanting to calculate (and perhaps greater context for the more general problem you are looking to solve).
(ii) Your question (with regard to CIs) is heavily dependent on the answer to the first.

The calculation of "number of hospitalizations" is very definition-driven. So in this regard there is no single method to do this that is "THE method". Furthermore, it would help to know if you're looking to (i) estimate this number from data you have, or (ii) simply to look-up how other people make these calculations.

Sources of variation for the definition of hospitalization rates include
(i) variation in ward
(ii) whether this was an admission vs a readdmission
(iii) whether and how readmissions are merged with the original admission.
- by proximity in time
- by underlying cause of admission
(iv) grouping related admissions to different hospitals or hospital systems

Going through these items in detail will help us get you to the right sources of information.






------------------------------
Glen Wright Colopy
DPhil Student
University of Oxford

Original Message:
Sent: 04-12-2019 13:44
From: David Ipe
Subject: Number of hospitalization per patient per month

Do you know of any literature that provides teh method for calculating # of hospitalization per patient per month and it confidence interval.​

------------------------------
David Ipe
Pharmacyclics
------------------------------

I would fit the standard and zero-inflated models Dr. Landsittel suggested and see which model provides the best fit. Here is a tutorial in R:

https://stats.idre.ucla.edu/r/dae/zip/

Alternatively:

https://support.sas.com/documentation/cdl/en/statug/63347/HTML/default/viewer.htm#statug_glimmix_sect028.htm

Also, as Glen Colopy alluded to, if you are observing patients at several sites then you might also consider adding random intercepts associated with sites.


------------------------------
Robert O'Brien
------------------------------

Original Message:
Sent: 04-15-2019 12:35
From: David Ipe
Subject: Number of hospitalization per patient per month

This is an prospective observational study and we are observing patients who receive treatment at enrollment into the study all the way till they receive the next treatment after the index treatment (i.e treatment at enrollment).
A patient may have one or more hospitalizations during this observation period and we are interested in calculating the number of hospitalizations per patient per month for the entire cohort of patients who were enrolled in this prospective observation study. Some of the patients may have zero number of hospitalizations.
Hope this clarifies.
Thanks
David

------------------------------
David Ipe
Pharmacyclics

Original Message:
Sent: 04-15-2019 11:32
From: Glen Colopy
Subject: Number of hospitalization per patient per month

Hi David,
It sounds like you're working on an interesting project!

I'd quickly second Prof. Landsittel's reply that
(i) It would help to have greater detail about the exact statistic you are wanting to calculate (and perhaps greater context for the more general problem you are looking to solve).
(ii) Your question (with regard to CIs) is heavily dependent on the answer to the first.

The calculation of "number of hospitalizations" is very definition-driven. So in this regard there is no single method to do this that is "THE method". Furthermore, it would help to know if you're looking to (i) estimate this number from data you have, or (ii) simply to look-up how other people make these calculations.

Sources of variation for the definition of hospitalization rates include
(i) variation in ward
(ii) whether this was an admission vs a readdmission
(iii) whether and how readmissions are merged with the original admission.
- by proximity in time
- by underlying cause of admission
(iv) grouping related admissions to different hospitals or hospital systems

Going through these items in detail will help us get you to the right sources of information.






------------------------------
Glen Wright Colopy
DPhil Student
University of Oxford

Original Message:
Sent: 04-12-2019 13:44
From: David Ipe
Subject: Number of hospitalization per patient per month

Do you know of any literature that provides teh method for calculating # of hospitalization per patient per month and it confidence interval.​

------------------------------
David Ipe
Pharmacyclics
------------------------------

​In number 4, you suggest that you are tracking the number of admissions per month over time.  Could you tell us why?  I have worked with time series hospital data for admissions and I find comparisons within group to be subject to regression to the mean effects.  I use a propensity score matching approach to compare the results of the study group to a matched sample.  I find that there is typically a peak during a crisis followed by a decline.  This peak seems to occur for both the study group and the matched sample.

------------------------------
Thomas Arnold
Population Health Data Scientist
CentraCare Health
------------------------------

Original Message:
Sent: 04-15-2019 12:35
From: David Ipe
Subject: Number of hospitalization per patient per month

This is an prospective observational study and we are observing patients who receive treatment at enrollment into the study all the way till they receive the next treatment after the index treatment (i.e treatment at enrollment).
A patient may have one or more hospitalizations during this observation period and we are interested in calculating the number of hospitalizations per patient per month for the entire cohort of patients who were enrolled in this prospective observation study. Some of the patients may have zero number of hospitalizations.
Hope this clarifies.
Thanks
David

------------------------------
David Ipe
Pharmacyclics

Original Message:
Sent: 04-15-2019 11:32
From: Glen Colopy
Subject: Number of hospitalization per patient per month

Hi David,
It sounds like you're working on an interesting project!

I'd quickly second Prof. Landsittel's reply that
(i) It would help to have greater detail about the exact statistic you are wanting to calculate (and perhaps greater context for the more general problem you are looking to solve).
(ii) Your question (with regard to CIs) is heavily dependent on the answer to the first.

The calculation of "number of hospitalizations" is very definition-driven. So in this regard there is no single method to do this that is "THE method". Furthermore, it would help to know if you're looking to (i) estimate this number from data you have, or (ii) simply to look-up how other people make these calculations.

Sources of variation for the definition of hospitalization rates include
(i) variation in ward
(ii) whether this was an admission vs a readdmission
(iii) whether and how readmissions are merged with the original admission.
- by proximity in time
- by underlying cause of admission
(iv) grouping related admissions to different hospitals or hospital systems

Going through these items in detail will help us get you to the right sources of information.






------------------------------
Glen Wright Colopy
DPhil Student
University of Oxford

Original Message:
Sent: 04-12-2019 13:44
From: David Ipe
Subject: Number of hospitalization per patient per month

Do you know of any literature that provides teh method for calculating # of hospitalization per patient per month and it confidence interval.​

------------------------------
David Ipe
Pharmacyclics
------------------------------