["model"] Census for hospital and icu are not inclusive of longer LOS for (respectively) icu and ventilator

### Summary

From Slack:

civicanalyst Today at 5:06 PM
> Hi all, When I run CHIME for one of our hospitals in PA using the assumption that 30% of hospitalizations will need ICU and Vent along with the assumption that LOS of Vent > LOS ICU, the census curves for Vent is above the ICU case. I understand from documentation that Vent should be a subset of ICU and maybe higher LOS is affecting that? Has this already been discussed here or any thoughts on why that might be happening?

philmiller
> I think I see an underlying mistake in how we’re calculating the census for inpatient relative to icu, and icu relative to vent - admissions categories are inclusive, but we’re not counting a longer stay in the hospital for the portion of patients who end up in critical care

### Additional details

### Suggested fix
Calculate a census of the exclusive values, and then sum them appropriately.