Allocating surgical numbers to groups

Hi, I hope you can help.

I am looking at allocating hospitals to the following groups, LHA( Low hospital activity) MHA(medium activity HHA(High hospital activity. For which I will then perform analysis on the costs associated with delivering that surgery, as we have costs associated with each hospital and procedure. What I am trying to prove that those hospitals with low activity either have high or low costs.

What I am not sure of is the allocation to groups. Could I use quartile Upper, Middle and Lower or is there a better formula to assign?

Thanks for your help in advance


Less is more. Stay pure. Stay poor.
How many hospitals do you have? How are you defining activity (e.g., census, number of beds, etc.)?
How many hospitals do you have? How are you defining activity (e.g., census, number of beds, etc.)?
hey, thanks for your quick response. 135 hospitals which is about 88% of the full estate and the activity is based on HRG code associated with the small group of procedures under that HRG code. I also would like to look at the length of stay for each based on costs, but for now just activity.


Less is more. Stay pure. Stay poor.
Yeah I suppose you could trichotomize hospitals based on creating tertiles, but it doesn't seem like you have a definition for activity so why group them at all. Why not just treat activity level as a continuous variable? In medicine everyone is always trying to group things, but actually just losing information along the way. Can you tell me a person with a BMI of 29.99 is actually different from a person with a BMI 30.01, on any given day their BMI values could be switched, but calling one overweight and the other normal is ludicrous.

It would help if you posted a histogram of HRG based on the sample of 135 hospitals, so we can understand its distribution.
It is clear from a surgical approach that surgeons performance is better the more they perform. So in some cases, it is classed that for certain surgical techniques that those doing more than 15 operations a year is deemed competent. I would agree there are lots of variables, for example, depends on who taught them and the patient groups and profile.

However when looking at costs, you could argue that hospitals that do more, have a better control of costs, however in this case, I believe the issue is the those which try to repair a patient and only do 1 of these a year should not be doing it and therefore should be funnelled to experts. Does that make sense? However, the reverse of this could also be true.

The spread is 1 to 47 operations across the units and they have a cost of £412 to £20878 which is massive.

Thanks for you help


TS Contributor
I can still see no reason why to create 3 artifical groups. You can analyse the association between
cost and the activity measure(s) without any grouping. Categorization can cause unnecessary problems .

Or, are there any pre-existing criteria e.g. from literature or from the practice in your field, which
can be used for the assessment of an activity level as "high" or "middle" or "low"?

With kind regards

@Karabiner You are right I don't want to create the groups unless I have too. However, we completed the analysis of the activity cost by looking at Mean. Median, Min, Max, STDev, range and the same for activity. However, the challenge is seeing if the above mean by 1 or 2+SD, is their costs any higher or lower than those that are -2SD from the mean? Is it more acceptable to look at the upper qtr, mid and lower, to allocate the groups maybe?

What would you do?