Param Revamp - Other Adult Cancer

[mmsuarezcosta]

Updating other adult cancer cancer parameters (removing module hardcoding and adding labels)

To Note:
Discrepancies between comments and values (this is the original code before I removed the hard-coding. The parameter values are that which were previously hard-coded (I ignored the commetns):

       # ----- SCHEDULE MAIN POLLING EVENTS -----
        # Schedule main polling event to happen immediately
        sim.schedule_event(OtherAdultCancerMainPollingEvent(self), sim.date + DateOffset(months=1))

            # Schedule a post-treatment check for 12 months:
            hs.schedule_hsi_event(
                hsi_event=HSI_OtherAdultCancer_PostTreatmentCheck(
                    module=self.module,
                    person_id=person_id,
                ),
                topen=self.sim.date + DateOffset(months=3),
                tclose=None,
                priority=0
            )

            # Schedule another HSI_OtherAdultCancer_PostTreatmentCheck event in one month
            hs.schedule_hsi_event(
                hsi_event=HSI_OtherAdultCancer_PostTreatmentCheck(
                    module=self.module,
                    person_id=person_id
                ),
                topen=self.sim.date + DateOffset(months=3),
                tclose=None,
                priority=0
            )

Parameters to Review labeling:

• All init_prop: Currently labeled as 'undetermined', but unsure if should be labeled as 'local'
• min_age_adult_cancer: given that Malawi is specific to Under 5 and Over 5 OPD, I believe this is local
• Rates and relative rates (all that begin with r, rr, rp): Unsure between 'local' and 'universal'. Have not been able to trace each back to a reference. I took the approach that for any that was a rate of progression (e.g. r_metastatic_local_ln) label as 'universal', any that was a rate of diagnosis and treatment, I labeled it as 'local'

Reference Notes:
NA

[marghe-molaro]

Updating other adult cancer cancer parameters (removing module hardcoding and adding labels)

To Note: Discrepancies between comments and values (this is the original code before I removed the hard-coding. The parameter values are that which were previously hard-coded (I ignored the commetns):

Yes, I agree with this approach. I have raised an issue #1681 to flag this recurrent problem. Would be good to wait until resolution of this issue before merging this (and other param revamp) PRs.

Parameters to Review labeling:

• All init_prop: Currently labeled as 'undetermined', but unsure if should be labeled as 'local'

If write-up doesn't explicitly state this was calibrated, policy for now is to assume it wasn't, and ask developers to review later.

• min_age_adult_cancer: given that Malawi is specific to Under 5 and Over 5 OPD, I believe this is local

I agree, different settings may have different guidelines with regards to age thresholds for care

• Rates and relative rates (all that begin with r, rr, rp): Unsure between 'local' and 'universal'. Have not been able to trace each back to a reference. I took the approach that for any that was a rate of progression (e.g. r_metastatic_local_ln) label as 'universal', any that was a rate of diagnosis and treatment, I labeled it as 'local'

I would have done the same. We can always ask developers to double check.

[mmsuarezcosta]

@marghe-molaro added back comment of code and comment discrepancy

[tbhallett]

indeed -- so we should correct the comment!

[mmsuarezcosta]

Issue is that we currently don't know if it should be 3 months (current parameter value) or 12 months (current code value). Plan was to leave comment as-is.

I have updated the comment to be the current parameter value, and I have added the original value from the comment in the (NOTE) section

[marghe-molaro]

According to @andrew-phillips-1 the issue is in the code, not the comment (see discussion in Issue #1681)

[mmsuarezcosta]

You're right, thanks for the clarification here @marghe-molaro

[tbhallett]

Thanks so much @mmsuarezcosta and @marghe-molaro for this.

I've had a look at just this one and we're definitely on the right track.

I've added a few comments already, but on the actual paramneters.csv I couldn't add line specific comments, so have done it here instead.

In a couple of places, I would say the parameters could be 'universal' (rather than 'local') as indicated below. Let me know if you disagree. But this kind of pattern I would expect to be quite common across the modules, so I thought worth raising.

parameter_name	value	param_label	prior_min	prior_max	prior_note	reference	TH comment
init_prop_early_other_adult_ca_symptom_other_adult_cancer_by_stage	[0.3,0.5,0.8]	local	[0.0,0.0,0.0]	[1.0,1.0,1.0]	vague prior		Agree
in_prop_other_adult_cancer_stage	[0.0002, 0.0002, 0.0004]	local	[0.0, 0.0, 0.0]	[1.0, 1.0, 1.0]	vague prior		Agree
init_prop_with_early_other_adult_ca_symptom_diagnosed_by_stage	[0.3,0.4,0.5]	local	[0.0,0.0,0.0]	[1.0,1.0,1.0]	vague prior		Agree
init_prop_treatment_status_other_adult_cancer	[0.1,0.1,0.1]	local	[0.0,0.0,0.0]	[1.0,1.0,1.0]	vague prior		Agree
init_prob_palliative_care	0.95	local	0	1	vague prior		Agree
r_site_confined_none	1.00E-05	local	0	1	vague prior		Agree
rr_site_confined_age3049	5	local	0.01	100	vague prior		Universal? The effects of co-factor on incidence will be informed by large studies and is not expected to be specific to local context neccesarily.
rr_site_confined_age5069	21	local	0.01	100	vague prior		Universal? The effects of co-factor on incidence will be informed by large studies and is not expected to be specific to local context neccesarily.
rr_site_confined_age70	60	local	0.01	100	vague prior		Universal? The effects of co-factor on incidence will be informed by large studies and is not expected to be specific to local context neccesarily.
rr_site_confined_hiv	4.13	local	0.01	100	vague prior		Universal? The effects of co-factor on incidence will be informed by large studies and is not expected to be specific to local context neccesarily.
r_local_ln_site_confined_other_adult_ca	0.15	universal	0	1	vague prior		Agree
rr_local_ln_other_adult_ca_undergone_curative_treatment	0.1	universal	0.01	100	vague prior		Agree
r_metastatic_local_ln	0.15	universal	0	1	vague prior		Agree
rr_metastatic_undergone_curative_treatment	0.2	universal	0.01	100	vague prior		Agree
r_death_other_adult_cancer	0.2	universal	0	1	vague prior		Agree
r_early_other_adult_ca_symptom_site_confined_other_adult_ca	0.02	local	0	1	vague prior		Agree
rr_early_other_adult_ca_symptom_local_ln_other_adult_ca	3	local	0.01	100	vague prior		Universal? The effects of co-factor on incidence will be informed by large studies and is not expected to be specific to local context neccesarily.
rr_early_other_adult_ca_symptom_metastatic_other_adult_ca	10	local	0.01	100	vague prior		Universal? The effects of co-factor on incidence will be informed by large studies and is not expected to be specific to local context neccesarily.
rp_other_adult_cancer_age3049	1.5	local	0.01	100	vague prior		Agree
rp_other_adult_cancer_age5069	2	local	0.01	100	vague prior		Agree
rp_other_adult_cancer_age70	4	local	0.01	100	vague prior		Agree
sensitivity_of_diagnostic_device_for_other_adult_cancer_with_other_adult_ca_site_confined	0.8	universal	0	1	vague prior		Agree
sensitivity_of_diagnostic_device_for_other_adult_cancer_with_other_adult_ca_local_ln	0.95	universal	0	1	vague prior		Agree
sensitivity_of_diagnostic_device_for_other_adult_cancer_with_other_adult_ca_metastatic	0.95	universal	0	1	vague prior		Agree
odds_ratio_health_seeking_in_adults_early_other_adult_ca_symptom	4	local	0.01	100	vague prior		Agree
post_treatment_check_months	3	local	1	12	vague prior		Agree
follow_up_appt_months	3	local	1	12	vague prior		Agree
palliative_care_interval_months	1	local	1	12	design decision		Agree
treatment_bed_days	5	local	1	30	vague prior		Agree
palliative_care_bed_days	15	local	1	50	vague prior		Agree
min_age_adult_cancer	5	local	0	18	vague prior		Agree
initial_polling_start_months	1	undetermined	1	12	design decision		Agree
initial_polling_start_months_palliative_care	1	undetermined	1	12	design decision		Agree
initial_polling_start_delay_weeks_palliative_care	1	undetermined	1	12	design decision		Agree
main_polling_event_frequency_months	1	undetermined	1	12	design decision		Agree

[mmsuarezcosta]

• rr_site_confined_age3049, rr_site_confined_age5069, rr_site_confined_agege70: Agreed that these are universal in principle. However, as I was going through the write-up, I deemed that these values were derived from a local source based on the following statement: "Jointly, the chosen parameter values produce the model outputs shown in Table 3. Informed by incidence rates of other_adult_cancers from Malawi cancer registry and other studies (Chasimpha et al, 2017), incidence of other_adult_cancers is assumed to be dependent on age. "

• rr_site_confined_hiv: Agree. Write-up states universal source

• rr_early_other_adult_ca_symptom_local_ln_other_adult_ca, rr_early_other_adult_ca_symptom_metastatic_other_adult_ca: Agree, and do not find reference in write-up to deem otherwise

[andrew-phillips-1]

I would think it was hard to regard any of these as universal since this is a mixture of all other adults cancers ?

[mmsuarezcosta]

I would think it was hard to regard any of these as universal since this is a mixture of all other adults cancers ?

That is a fair point. At the same time, the way we have defined 'universal' is that the variable should be agnostic of local differences. Thus, anything that has to with test sensitivity/ specificity, disease progression, or death rates are defined as 'universal'. I would argue to keep the values defined as universal above as-is.

Would be good to also have @marghe-molaro input on this.

[andrew-phillips-1]

Happy for you to stick with "universal" for the reasons you state.