Treatment of SAM extension only covers some MAM children

CHANGELOG.md

@@ -1,8 +1,11 @@
 
 ### Changelog
 
+#### [2.0.10]
+- If treatment of SAM is expanded to treatment of MAM, the MAM component only covers former SAM children
+
+
 #### [2.0.9]
-- If treatment of SAM is expended to treatment of MAM, the MAM component only covers former SAM children
 - Reduce threshold for filtering programs in complex objective function optimizations
 
 

nutrition/model.py

@@ -175,8 +175,6 @@ def _track(self):
     @translate
     def _set_pop_probs(self, year):
         init_cov = self.prog_info.get_ann_covs(year - 1)
-        # Treatment of SAM must be relative to initial year to avoid MAM extension issues. Other coverages are updated iteratively
-        #init_cov[_("Treatment of SAM")] = self.prog_info.get_ann_covs(0)[_("Treatment of SAM")]
         prog_areas = self.prog_info.prog_areas
         for pop in self.pops:
             pop.previousCov = init_cov
@@ -376,12 +374,23 @@ def _update_dists(self, population):
 
     @translate
     def _wasting_trans(self, age_group):
-        """Calculates the transitions between MAM and SAM categories
-        Currently only accounting for the movement from sam to mam, because the other direction is complicated"""
-        numsam = age_group.num_risk(_("SAM"))
-        nummam = age_group.num_risk(_("MAM"))
-        # If the denominator is 0.0 or close, set update to 1 (no change).
-        age_group.fromSAMtoMAMupdate[_("MAM")] = 1 + sc.safedivide(((1.0 - age_group.wastingTreatmentUpdate[_("SAM")]) * numsam), nummam, default=0.0)
+        """Calculates the transitions INTO MAM category
+        For wasting treatment intervention, children leaving SAM are accounted for in programs.py.
+        If MAM extension is selected, then children leaving MAM are also accounted for in programs.py.
+        This function accounts for the inflow to MAM (whether or not MAM extension is selected)."""
+        if len(age_group.data.or_wasting_prog) == 1: # if treatment of MAM extension is not selected
+            numsam = age_group.num_risk(_("SAM"))
+            nummam = age_group.num_risk(_("MAM")) * age_group.wastingTreatmentUpdate[_("MAM")]
+            age_group.fromSAMtoMAMupdate[_("MAM")] = 1 + sc.safedivide(((1.0 - age_group.wastingTreatmentUpdate[_("SAM")]) * numsam), nummam, default=0.0)
+        else:
+            age_group.fromSAMtoMAMupdate[_("MAM")] = 1
+            # TESTING differences for increasing vs decreasing coverage
+            # if age_group.wastingTreatmentUpdate[_("SAM")] <= 1: # if treatment program coverage is increasing
+            #     age_group.fromSAMtoMAMupdate[_("MAM")] = 1 #+ sc.safedivide(((1.0 - age_group.wastingTreatmentUpdate[_("SAM")]) * numsam), nummam, default=0.0)
+            # else:
+            #     #numsam = age_group.num_risk(_("SAM")) / age_group.wastingTreatmentUpdate[_("SAM")]
+            #     #nummam = age_group.num_risk(_("MAM")) / age_group.wastingTreatmentUpdate[_("MAM")]
+            #     age_group.fromSAMtoMAMupdate[_("MAM")] = 1# + sc.safedivide(((1.0 - age_group.wastingTreatmentUpdate[_("SAM")]) * numsam), nummam, default=0.0)
 
     @translate
     def _dia_indirect_effects(self, age_group):

nutrition/populations.py

@@ -642,7 +642,12 @@ def _set_prob_wasted(self, prog_areas):
                         OR = self.data.or_wasting_prog[wastingCat][program][age]
                     except Exception:  # if, for eg, MAM doesn't have Treatment of SAM key (odict doesn't have keyerror)
                         OR = 1
-                    fracCovered = self.previousCov[program]
+                    if wastingCat ==_("MAM"):
+                        # If SAM treatment is extended to MAM, then the coverage refers to continuing treatment
+                        # for SAM children (from MAM to mild), who are only a fraction of MAM children
+                        fracCovered = min(1, self.previousCov[program] * age_group.frac_wasted(_('SAM')) / age_group.frac_wasted(_('MAM')))
+                    else:
+                        fracCovered = self.previousCov[program]
                     pn, pc = solve_quad(OR, fracCovered, fracThisCatAge)
                     age_group.probConditionalCoverage[wastingCat][program] = {}
                     age_group.probConditionalCoverage[wastingCat][program]["covered"] = pc

nutrition/programs.py

@@ -293,14 +293,15 @@ def _wasting_prev_update(self, age_group):
             oldProb = age_group.frac_wasted(wastingCat)
             probWastedIfCovered = age_group.probConditionalCoverage[wastingCat][self.name]["covered"]
             probWastedIfNotCovered = age_group.probConditionalCoverage[wastingCat][self.name]["not covered"]
-            # if self.name == _("Treatment of SAM") and wastingCat == _('MAM'):
-            #     # If SAM treatment is extended to MAM, then the coverage refers to SAM children, who are only a fraction of MAM children
-            #     newcov = min(1, self.annual_unrestr_cov[self.year] * (age_group.frac_wasted(_('SAM')) / age_group.frac_wasted(_('MAM'))))
-            #     newProb = get_new_prob(newcov, probWastedIfCovered, probWastedIfNotCovered)
-            #     print(self.year, age_group.age, newcov, probWastedIfCovered, newProb, oldProb)
-            # else:
-            newcov = self.annual_unrestr_cov[self.year]
-            newProb = get_new_prob(newcov, probWastedIfCovered, probWastedIfNotCovered)
+            if self.name == _("Treatment of SAM") and wastingCat == _('MAM'):
+                # If SAM treatment is extended to MAM, then the coverage refers to SAM children, who are only a fraction of MAM children
+                newcov = min(1, self.annual_unrestr_cov[self.year] * age_group.frac_wasted(_('SAM')) / age_group.frac_wasted(_('MAM')))
+                newProb = get_new_prob(newcov, probWastedIfCovered, probWastedIfNotCovered)
+                print(self.year, age_group.age, newcov, probWastedIfCovered, newProb, oldProb,
+                      sc.safedivide(oldProb - newProb, oldProb, default=0.0), probWastedIfCovered - probWastedIfNotCovered)
+            else:
+                newcov = self.annual_unrestr_cov[self.year]
+                newProb = get_new_prob(newcov, probWastedIfCovered, probWastedIfNotCovered)
             reduction = sc.safedivide(oldProb - newProb, oldProb, default=0.0)  # If the denominator is 0.0 or close, set reduction to zero (no change)
             update[wastingCat] = 1 - reduction
         return update

nutrition/version.py

@@ -1,5 +1,5 @@
 import sciris as sc
 
-version = "2.0.9"
-versiondate = "2023-03-10"
+version = "2.0.10"
+versiondate = "2023-06-30"
 gitinfo = sc.gitinfo(__file__)