Analyzing and visualizing patient flow in an emergency department

my team has a 6-month dataset of individual patients in the emergency departments of three hospitals. Variables include demographics (sex, gender), time of arrival and departure (and thus length-of-stay), and which diagnoses the patient received. In addition, we have registered the triage category, which reflects the degree of emergency (red is most urgent, followed by orange, yellow and green).

We have performed simple analyses such as median length-of-stay at the three hospitals, categorized by groups of diagnoses, and categorized by triage category.

We now want to analyze variations of the length-of-stay. Our questions include:
- [Graphical representation] How does the length-of-stay vary during the day and week?
- [Graphical representation and/or table] How does the number of patients that are in the emergency department at the same time, affect length-of-stay?
- [Regression?] What are the most important factors contributing to a long length-of-stay?

I have attached a few graphs from a similar project. How can we make graphs such as these?
The text is in Norwegian, so here is an explanation: The first graph shows median length-of-stay in hours (y-axis) for the four triage categories, on each day of the week (x-axis). The second graph shows number of patients (y-axis) on each day of the week (x-axis), and the colors represent percentiles for the study period.

We use SPSS. I acknowledge that this thread contains many questions, but I am thankful for all help.



No cake for spunky
I don't think six months is enough data to make such an analysis. If seasonality exists or if the time was unusual you are not going to learn much from such a short period of time. Unless those are the only six months you are interested in.

Unfortunately I do not know those charts. Have you contacted SPSS or looked at their documentation on charts - I always do this when I am trying to build charts. They will have an on line help line.
Thank you for the quick reply.
Six months: We have more than 16,000 subjects. Don’t your think that’s enough? For example to evaluate whether more simultaneous patients in the emergency department equals longer length-of-stay.

An additional, related question: What is a good way to describe number of patients at given time points? Example: Patient A is in the emergency department 10:00-13:59; Patient B 11:00-12:59; Patient C 12:00-14:59. That means there is 1 patient at 10:00, 2 patients at 11:00, 3 at 12:00, 2 at 13:00, 1 at 14:00. I struggle to find a good way to describe this in a variable.

Charts: Contacting SPSS was a clever tip. Will do!
LOS in the ED / the time it takes in the ED. Time and date is recorded for patient arrival to the ED, and departure (to another department or out of the hospital).


No cake for spunky
"Six months: We have more than 16,000 subjects. Don’t your think that’s enough? For example to evaluate whether more simultaneous patients in the emergency department equals longer length-of-stay."

What if some unusual event, like COVID 19, occurred in those six months and distorted the results. What if you measure DEC through May and summer has a different reality. Its not the number of people that is the issue, its seasonality and the possibility of special causes.
Thank you for the reply.
That's certainly an important point. However, I think that's a point belonging to the discussion of our results, not whether or not a calculation is possible.

We are running preliminary analyses on these six months. However, the methods we employ will be used as a template for a future, prospective data collection.


No cake for spunky
The calculation is possible I think based on what you have suggested. If you are asking about the charts. I think your first question can be done by simple descriptive or showing overall variation. I think your second and 3rd question can be done through regression.

I don't know the charts in question. SPSS is pretty straightforward in its charts from memory (I have not used it in many years). If you have trouble with it you should be able to find information on the specific graphs on line or contact SPSS.