I work at a hospital in a coastal area that is prone to flooding. The hospital has a long list of tasks that are performed in the event of an emergency, and a large (over 1,000) set of organizational units that those tasks are doled out to.
They would like to assess their ability to respond to emergencies with a "simulation", and I have some doubts about some of their methodology.
For instance, one of the tasks is to build a sandbag fortification around an emergency generator. There are 50 different units in the hospital that are supposed to be prepared to carry out this task within four hours of being notified. Their plan is to randomly select three of them and see if they can accomplish the task or not.
These units are very "heterogeneous". Some are significantly larger than others, some are demographically skewed younger, or more male (which I suspect could be an issue when you're talking filling and moving sandbags), etc.
My concern is that with such a small sample, and a binary task outcome, can you really generalize from these results?
If all three organizations "pass" can you really say that the hospital as a whole is well prepared on this task? What if you happened to select the three largest, most male units? Or what if only two of the three pass, or one? What does that say about the capability of the hospital as a whole?
I've been arguing that they should a) use a larger sample size and b) measure the time taken taken to complete the task to satisfaction. Then they could form a confidence interval on the time to complete.
If they stick to the current methodology, can a general observation be made? If so, via what method? And if not, is there a "name" for this kind of deficiency?
Thanks!
As to your question: There is insufficient information to be able to answer. It depends on a) How many of the units will be on duty at any given time b) How many of the units have to build sandbags in order for enough sandbags to be built. And probably some other factors as well.
– Peter Flom Feb 06 '24 at 17:26