cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Choose Language Hide Translation Bar
JP27
Level I

Newbie - What am I doing (forecasting hospital admission/Time Series analysis)

Newbie here with little experience in coding (medical background). 

 

I was given this database which contains all hospital admissions during a 1 year period.

Each row represents an individual admission, and columns contain variables like date of admission (YY/MM/DD hh:mm) or source of admission (home, transfer from other hospital, etc). 

JP27_1-1683222803197.png

This is a very simplified version of how my table looks like. Actual table contains thousands of admissions. Admission date is in ascending order.

 

I was instructed to develop a forecasting model to see how many admissions we can expect in the following month/year based on this data. I did a bit of research and consider that a Time Series analysis could work for our purpose. It should be something simple: to forecast future daily admissions based on current daily admissions. 

 

The only way I have been able to get some output was by using by admission date variable in the Y axis, and I am getting something like this:

JP27_0-1683222583533.png

 

Instead, I am hoping to get something like this:

JP27_2-1683223036043.png

 

I know I am probably missing some super basic concepts. I read and did the exercises in the JMP book but still I can't solve this. Thank you in advance!

 

3 REPLIES 3
statman
Super User

Re: Newbie - What am I doing (forecasting hospital admission/Time Series analysis)

Here are my thoughts:

1. History does not predict the future.

2. You want a model to predict the future, you need to study the admissions process to get an understanding of causal factors that affect admissions.  You might also question the admissions data.  How accurate is it?  Are the times/dates correct?  Are there potential opportunities to distort the data (for legal or financial reasons)?

3. I would use control charts to first answer the question "Is the admissions process consistent and stable?"  If so, you may extend those control limits as a "prediction" of future admissions given the future conditions are appropriately represented in the data set you currently have (for example, what if we are hit with another pandemic or natural disaster and this  is not represented in the current 1 year data?).  If not, you should seek to understand why.

"All models are wrong, some are useful" G.E.P. Box
FN
FN
Level VI

Re: Newbie - What am I doing (forecasting hospital admission/Time Series analysis)

I would look into the Time Series Forecasting platform.

https://www.jmp.com/en_gb/learning-library/topics/time-series/time-series-forecasting.html

 

With a good read of this great and free book:

https://otexts.com/fpp2/

 

 

 

 

peng_liu
Staff

Re: Newbie - What am I doing (forecasting hospital admission/Time Series analysis)

I cannot see the y-axis label in your targeted result due to resolution. But I believe that is not in your data. So a hint is can you get the data into the shape that is required for your targeted plot?