Integrating weather events into TLO#1604
Integrating weather events into TLO#1604RachelMurray-Watson wants to merge 328 commits intomasterfrom
Conversation
marghe-molaro
left a comment
marghe-molaro
left a comment
There was a problem hiding this comment.
I had a quick look at the changes to the HealthSystem, please let me know if any of my comments are unclear!
src/tlo/methods/healthsystem.py
Outdated
| # And for each indiviudal level event, check to see if there are projected disruptions due to precipitation. | ||
| if self.module.services_affected_precip != 'none' and year > 2025: | ||
| for item in list_of_individual_hsi_event_tuples_due_today_that_have_essential_equipment: | ||
| facility_used = self.sim.population.props.at[item.hsi_event.target, 'level_1a'] |
There was a problem hiding this comment.
Why is only 'level_1a' considered here? The facility at which HSI is being delivered might not necessarily be at this level
There was a problem hiding this comment.
Hang-over from previous code; thanks for spotting!
src/tlo/methods/healthsystem.py
Outdated
| if self.module.services_affected_precip != 'none' and year > 2025: | ||
| for item in list_of_individual_hsi_event_tuples_due_today_that_have_essential_equipment: | ||
| facility_used = self.sim.population.props.at[item.hsi_event.target, 'level_1a'] | ||
| if facility_used in self.module.parameters['projected_precip_disruptions']['Facility_ID'].values: |
There was a problem hiding this comment.
If I understand correctly, 'Facility_ID' here refers to real facilities. Since 'Facility_ID' is already used in the model to indicate modelled/aggregated facilities, I would use a different name if possible - e.g. RealFacility_ID? - to avoid confusion
There was a problem hiding this comment.
It might also be clearer to separate the mapping of the modelled and real facilities from the climate considerations, so have dictionaries in the health system ready to be invoked where:
Real_to_Modelled_Facility_Mapping = {RealFacility_ID : Facility_ID}
and
Modelled_to_Real_Facility_Mapping = {Facility_ID : [RealFacility_ID, RealFacility_ID, ...]}
src/tlo/methods/healthsystem.py
Outdated
|
|
||
| # Try to run the list of individual-level events that have their essential equipment | ||
| # And for each indiviudal level event, check to see if there are projected disruptions due to precipitation. | ||
| if self.module.services_affected_precip != 'none' and year > 2025: |
There was a problem hiding this comment.
Also if not None?
The type of service provided by the HSI under consideration doesn't seem to be accounted for currently (meaning only the 'none' or 'all' services being disrupted is currently under consideration, so I would remove the 'anc' option for now
src/tlo/methods/healthsystem.py
Outdated
| 'service'] == self.module.services_affected_precip), | ||
| 'disruption' | ||
| ] | ||
| if len(prob_disruption) > 0: |
There was a problem hiding this comment.
I'm a bit confused, why is the prob_distribution not a single value?
There was a problem hiding this comment.
So am I, unsure why I left that there!
src/tlo/methods/healthsystem.py
Outdated
| assert isinstance(prob_disruption, (int, float)), "prob_disruption must be an int or float" | ||
|
|
||
| if np.random.binomial(1, prob_disruption) == 1: # success is delayed appointment | ||
| list_of_individual_hsi_event_tuples_due_today_that_have_essential_equipment.remove( |
There was a problem hiding this comment.
The coin toss to check precipitation disruption should be done before adding the item to the list_of_individual_hsi_event_tuples_due_today_that_have_essential_equipment , rather than adding+removing
Call 'equipment is available' (cond A) and 'event is disrupted by precipitation (cond B), then:
If cond A=False:
call never_ran()
elif A=True:
check cond B
if B=False:
add to list_of_individual_hsi_event_tuples_due_today_that_have_essential_equipment
else:
invoke did_not_run(), and add to hold over list
In this respect, would be good to rename list_of_individual_hsi_event_tuples_due_today_that_have_essential_equipment with something more generic, like list_of_individual_hsi_event_tuples_meeting_all_conditions, or something along those lines
src/tlo/methods/healthsystem.py
Outdated
| if np.random.binomial(1, prob_disruption) == 1: # success is delayed appointment | ||
| list_of_individual_hsi_event_tuples_due_today_that_have_essential_equipment.remove( | ||
| item) # Remove item from list? Should I create a seperate list? | ||
| self.module.call_and_record_never_ran_hsi_event(hsi_event=item.hsi_event, |
There was a problem hiding this comment.
If the event has been postponed, rather than completely cancelled, we should not be calling "call_and_record_never_ran_hsi_event" on it, rather we should be adding it to the "hold_over" queue, but first calling "did_not_run()" fnc on it
There was a problem hiding this comment.
Per conversation 27th March 2025, decided to "just" cancel all appointments pending further nuancing of healthcare seeking behaviour.
src/tlo/methods/healthsystem.py
Outdated
| assert climate_model_ensemble_model in ('lowest', 'mean', 'highest') | ||
| self.climate_model_ensemble_model = climate_model_ensemble_model | ||
|
|
||
| assert services_affected_precip in (None, 'none', 'ANC', 'all') |
There was a problem hiding this comment.
currently only 'none' or 'all' is considered, so remove ANC option for now
Also edited the report_live_years_lost
And extra slice for indexing at end of year
- As a placeholder, if the disruption probability is high, cancelled. - If not, and there is perfect healthcare seeking, delay is 1 month (to be varied). - Delay is also 1 month for imperfect healthcare seeking, hopefully to be nuanced
Added option for mode 2 analysis
added in vmin and vmax again for clarity
…arameter scans, all with mode 2)
…arameter scans, all with mode 2)
2 participants
This PR aims to integrate weather-mediated disruptions into the TLO model. As the disruptions are mediated at the level of a facility, each individual in the model must be "assigned" to facilities of different levels within their region of interest.
Assumptions: