Param Revamp - Postnatal Supervisor

[mmsuarezcosta]

Parameters

Additional comment deleted (some params had values in addt notes not used)

• weekly_prob_pre_eclampsia_pn: [0.0016, 0.0016]
• probs_for_mgh_matrix_pn: [[0.884, 0.066, 0.05, 0, 0], [0.884, 0.066, 0.05, 0, 0]]
• probs_for_mpe_matrix_pn: [[0, 0, 0.89, 0.11, 0], [0, 0, 0.89, 0.11, 0]]
• rr_secondary_pph_endometritis : Delete? (Did not delete this parameter, it is used in postnatal_supervisor_lm.py)

Undetermined, Design Decision

• emergency_postnatal_care_reset_week (week == 6): Params reset at week 6 for women who are experiencing severe complications. Open discussion to a) revert to keeping this variable hard-coded as it is "by definition", b) update variable name if there is a specific logic for it that is more descriptive than 'reset', or c) accept new change as-is
• postnatal_reset_week (week == 8): Same discussion and options as above
• Kept variable week_5_postnatal_neonates as-is because seemed to be best categorized as option a in options above

Undetermined, Calibration

In some instances, for Parameter X the column Source and/or relevant calculation, the text will state See (name of Parameter Y). In these cases, when Parameter Y is derived via calibration, my understanding is that the Parameter X is also derived from calibration. Is this interpretation correct? @joehcollins

Examples include:

• prob_type_of_anaemia_pn: The write-up references baseline_prob_anaemia_per_month, which is defined as being derived by calibration. The definition of this parameter itself states that it is calculated as "the proportion of pregnant women with mild/moderate/severe anaemia divided by the total proportion of pregnant women with any anaemia".
• Same logic for parameter cfr_secondary_postpartum_haemorrhage. cfr_secondary_postpartum_haemorrhage references the calibrated parameter prob_ectopic_pregnancy_death. This calibrated parameter has the following description:

"As such untreated case fatality parameters have been estimated
through the process of calibration to ensure that the model
replicates both the assumed MMR and the proportion of deaths by
cause.

• Same logic for cfr_eclampsia
• cfr_severe_pre_eclampsia: Cannot find parameter in write-up, but assume that it was calibrated in same process as cfr_eclampsia

Local v. Universal

It is a little grey differentiating local v. universal parameters in this module. Will be great to have developer input here. I am aiming to capture the essence of

• 'truly universal': something that is to be true in MCH regardless of location
• 'local': can vary based on location

When unsure, I have erred on the side of 'local'. I am documenting the params that are most important to review below

Local

• prob_early_onset_neonatal_sepsis_week_1, prob_late_onset_neonatal_sepsis,
• All params with suffix _window_days, I am labeling as a local variable because it is based on local context how long appt can be scheduled for.
• rr_anaemia_maternal_malaria: sources based in Africa

Universal

• rr_gest_htn_obesity
• Most rr in this module I have defined as local except for: those with prefix rr_eons or treatment_effect
  , rr_gest_htn_obesity, rr_pre_eclampsia_obesity, rr_pre_eclampsia_diabetes_mellitus

Based on assumptions (documenting here for future reference)

In the write-up, it states that there was no/limited data for these values, so they are assumptions

• prob_htn_resolves, also taking this to be local bc I am assuming that there are local elements that are affecting this.
• prob_monthly_death_severe_htn
• weekly_prob_death_severe_gest_htn