Long-term hospital capacity planning model

[Ashish_Kumar]

Hello,

My colleagues and I are defining a long-term general hospital capacity planning model. It will have a time horizon of about 70 years, and will essentially model population ageing and its impact on needs for services in different hospital sub-systems (emergency department, inpatient, outpatients).

A population view will be the first of our Vensim views. We will use data on age-specific hospital utilisation to predict changes as the population structure and prevalence of health problems changes.

The logical unit of time for the model is years.

However, patient flows through emergency and outpatient departments happen in hours (e.g., 3-6 hours) and flows through inpatient areas in days (often 2-6 days).

I am looking for thoughts on how a stock and flow model should be designed in this context. One idea is that we have a model with a 70 year time horizon and yearly units. Within each year, the variables representing the usage of the various departments are auxiliary variables with the appropriate units. We would apply queueing model concepts in the year-on-year sub-system views. E.g., in the emegrency department, 10 arrivals per hour during a year with an average Length of Stay of 4 hours would mean 40 patients to be accomodated on avearge (of course we would need to buffer for variation). The disadvantage with this approach of using auxiliary variables is that we would lose the very useful insight from a stock-and-flow construction of the syb-system.

We could build two separate models, one with a long-term (70 years, unit of time years) view and the other with a short-term view (One month, unit of time hours).

I'll be grateful for ideas and pointers on what can be done in this kind of case, where the main aim is to have a long-term, high-level forecast, but the short-term dynamics are also important.

[LAUJJL]

Hi

I would start by building first a model for the short time that will use eventually queues, because once this model is built it will give insights for the longer time model. if may be possible later on to build a model for the short and long time included, but I have a doubt about it. It will be difficult to understand for the clients and understanding is mandatory from my point of view, being myself a modeler and a client at the same time. The short time model is too easier to build. Looking 70 years ahead is a difficult challenge.

For the first model, I would use a top down approach, starting first with a very simple model, to help the understanding of the model. I would too spend time, analyzing the model after each top down refinement, and even eventually trying to implement intermediate top down models to make practical experiments.

Regards.

JJ

[Ashish_Kumar]

Hi JJ,

Thanks for sharing your advice. True, looking that far ahead is challenging.

Best regards,
Ashish