CriticalHealthProfile now anchors ward admissions at severeness_onset (now on parity with SevereHealthProfile)

Author: Julian-Patzner

data/DefaultConf.toml

@@ -142,16 +142,16 @@
                     icu_admission_to_ventilation_admission = 0
                     ventilation_admission_to_ventilation_discharge = 0
                     ventilation_discharge_to_icu_discharge = 0
-                    # hospital admission delay (after critical onset)
-                    [Pathogens.Covid19.progressions.Critical.health.critical_onset_to_hospital_admission]
+                    # ward admission delay (after severeness onset), unless the ICU claims the patient first
+                    [Pathogens.Covid19.progressions.Critical.health.severeness_onset_to_hospital_admission]
                         distribution = "Poisson"
                         parameters = [1]
                     # ward stay length when not admitted to ICU (after hospital admission)
                     [Pathogens.Covid19.progressions.Critical.health.hospital_admission_to_hospital_discharge]
                         distribution = "Poisson"
                         parameters = [10]
-                    # ICU admission delay (after hospital admission)
-                    [Pathogens.Covid19.progressions.Critical.health.hospital_admission_to_icu_admission]
+                    # ICU admission delay (after critical onset)
+                    [Pathogens.Covid19.progressions.Critical.health.critical_onset_to_icu_admission]
                         distribution = "Poisson"
                         parameters = [1]
                     # ICU stay length when not ventilated (after ICU admission)

docs/src/config-files.md

@@ -132,13 +132,13 @@ If you want to set up a custom config file, you can copy this one into your own
                     icu_admission_to_ventilation_admission = 0
                     ventilation_admission_to_ventilation_discharge = 0
                     ventilation_discharge_to_icu_discharge = 0
-                    [Pathogens.Covid19.progressions.Critical.health.critical_onset_to_hospital_admission]
+                    [Pathogens.Covid19.progressions.Critical.health.severeness_onset_to_hospital_admission]
                         distribution = "Poisson"
                         parameters = [1]
                     [Pathogens.Covid19.progressions.Critical.health.hospital_admission_to_hospital_discharge]
                         distribution = "Poisson"
                         parameters = [10]
-                    [Pathogens.Covid19.progressions.Critical.health.hospital_admission_to_icu_admission]
+                    [Pathogens.Covid19.progressions.Critical.health.critical_onset_to_icu_admission]
                         distribution = "Poisson"
                         parameters = [1]
                     [Pathogens.Covid19.progressions.Critical.health.icu_admission_to_icu_discharge]

src/pathogen/health_progression/health_profiles.jl

@@ -51,16 +51,18 @@ end
 Health profile for an infection whose peak tier is `critical`: a hospital admission that can
 escalate to ICU and ventilation, plus mortality risk. Each step's probability is conditional
 on the step below (`hospital_to_icu_probability` = P(ICU | hospitalized), `icu_to_ventilation_probability` =
-P(ventilation | ICU)); discharges chain inward-out so the stays nest by construction. Timings
-are anchored at the infection's `critical_onset`. Care and mortality are computed independently
-here and reconciled once, downstream, by `compute_health!`.
+P(ventilation | ICU)); discharges chain inward-out so the stays nest by construction. Care and
+mortality are computed independently here and reconciled once, downstream, by `compute_health!`.
+
+The ward is anchored at `severeness_onset`, the ICU at `critical_onset`. If the ICU admission comes
+first, the ward admission is moved to it.
 
 # Parameters
 - `hospital_probability::Real`: Hospital admission probability (`0.0` by default).
-- `critical_onset_to_hospital_admission::Union{Distribution, Real}`: Admission delay after critical onset.
+- `severeness_onset_to_hospital_admission::Union{Distribution, Real}`: Admission delay after severeness onset, unless the ICU admission comes first.
 - `hospital_admission_to_hospital_discharge::Union{Distribution, Real}`: Ward stay length when the patient does not enter the ICU.
 - `hospital_to_icu_probability::Real`: ICU probability for a hospitalized patient (`0.0` by default).
-- `hospital_admission_to_icu_admission::Union{Distribution, Real}`: Delay from hospital to ICU admission.
+- `critical_onset_to_icu_admission::Union{Distribution, Real}`: ICU admission delay after critical onset.
 - `icu_admission_to_icu_discharge::Union{Distribution, Real}`: ICU stay length when the patient is not ventilated.
 - `icu_to_ventilation_probability::Real`: Ventilation probability for an ICU patient (`0.0` by default).
 - `icu_admission_to_ventilation_admission::Union{Distribution, Real}`: Delay from ICU to ventilation admission.
@@ -72,10 +74,10 @@ here and reconciled once, downstream, by `compute_health!`.
 """
 struct CriticalHealthProfile <: HealthProfile
     hospital_probability::Real
-    critical_onset_to_hospital_admission::Union{Distribution, Real}
+    severeness_onset_to_hospital_admission::Union{Distribution, Real}
     hospital_admission_to_hospital_discharge::Union{Distribution, Real}
     hospital_to_icu_probability::Real
-    hospital_admission_to_icu_admission::Union{Distribution, Real}
+    critical_onset_to_icu_admission::Union{Distribution, Real}
     icu_admission_to_icu_discharge::Union{Distribution, Real}
     icu_to_ventilation_probability::Real
     icu_admission_to_ventilation_admission::Union{Distribution, Real}
@@ -87,10 +89,10 @@ struct CriticalHealthProfile <: HealthProfile
 
     function CriticalHealthProfile(;
         hospital_probability = 0.0,
-        critical_onset_to_hospital_admission = 0,
+        severeness_onset_to_hospital_admission = 0,
         hospital_admission_to_hospital_discharge = 0,
         hospital_to_icu_probability = 0.0,
-        hospital_admission_to_icu_admission = 0,
+        critical_onset_to_icu_admission = 0,
         icu_admission_to_icu_discharge = 0,
         icu_to_ventilation_probability = 0.0,
         icu_admission_to_ventilation_admission = 0,
@@ -107,9 +109,9 @@ struct CriticalHealthProfile <: HealthProfile
             0.0 <= p <= 1.0 || throw(ArgumentError("$nm must be between 0 and 1 (got $p)."))
         end
 
-        return new(hospital_probability, critical_onset_to_hospital_admission,
+        return new(hospital_probability, severeness_onset_to_hospital_admission,
             hospital_admission_to_hospital_discharge, hospital_to_icu_probability,
-            hospital_admission_to_icu_admission, icu_admission_to_icu_discharge,
+            critical_onset_to_icu_admission, icu_admission_to_icu_discharge,
             icu_to_ventilation_probability, icu_admission_to_ventilation_admission,
             ventilation_admission_to_ventilation_discharge, ventilation_discharge_to_icu_discharge,
             icu_discharge_to_hospital_discharge, death_probability, critical_onset_to_death)
@@ -119,9 +121,9 @@ end
 """
     calculate_health_profile(cc::CriticalHealthProfile, individual::Individual, infection::InfectionState, rng::Xoshiro)
 
-Care and mortality contribution of a single `critical`-peak infection, anchored at
-`infection.critical_onset`. Hospital, ICU, and ventilation escalate via nested Bernoulli draws;
-death is drawn independently of them.
+Care and mortality contribution of a single `critical`-peak infection, the ward anchored at
+`infection.severeness_onset` and the ICU at `infection.critical_onset`. Hospital, ICU, and
+ventilation escalate via nested Bernoulli draws; death is drawn independently of them.
 """
 function calculate_health_profile(cc::CriticalHealthProfile, individual::Individual, infection::InfectionState, rng::Xoshiro)
     hospital_admission::Int16 = Int16(-1)
@@ -133,9 +135,11 @@ function calculate_health_profile(cc::CriticalHealthProfile, individual::Individ
     death::Int16 = Int16(-1)
 
     if gems_rand(rng) <= cc.hospital_probability
-        hospital_admission = rand_round(infection.critical_onset + _rand_val(cc.critical_onset_to_hospital_admission, rng), rng)
+        hospital_admission = rand_round(infection.severeness_onset + _rand_val(cc.severeness_onset_to_hospital_admission, rng), rng)
         if gems_rand(rng) <= cc.hospital_to_icu_probability
-            icu_admission = rand_round(hospital_admission + _rand_val(cc.hospital_admission_to_icu_admission, rng), rng)
+            icu_admission = rand_round(infection.critical_onset + _rand_val(cc.critical_onset_to_icu_admission, rng), rng)
+            # the ward admission cannot come after the ICU admission
+            hospital_admission = min(hospital_admission, icu_admission)
             if gems_rand(rng) <= cc.icu_to_ventilation_probability
                 ventilation_admission = rand_round(icu_admission + _rand_val(cc.icu_admission_to_ventilation_admission, rng), rng)
                 ventilation_discharge = rand_round(ventilation_admission + _rand_val(cc.ventilation_admission_to_ventilation_discharge, rng), rng)

src/pathogen/legacy_progressions/legacy_config.jl

@@ -12,11 +12,11 @@
     _is_legacy_critical(crit)
 
 `true` if a `Critical` progression config table uses the pre-decoupling health format, i.e. carries
-`icu_admission_to_death` or `severeness_onset_to_hospital_admission`. A modern `Critical` (new disease
+`icu_admission_to_death` or `hospital_discharge_to_severeness_offset`. A modern `Critical` (new disease
 tier, optionally with embedded care) carries neither.
 """
 _is_legacy_critical(crit) = crit isa Dict &&
-    (haskey(crit, "icu_admission_to_death") || haskey(crit, "severeness_onset_to_hospital_admission"))
+    (haskey(crit, "icu_admission_to_death") || haskey(crit, "hospital_discharge_to_severeness_offset"))
 
 """
     _normalize_legacy_pathogen!(params::Dict)

test/healthprogressiontest.jl

@@ -126,22 +126,22 @@ end
             critical_offset = Int16(15), severeness_offset = Int16(16), recovery = Int16(20))
         inf_crit = InfectionState(Int8(1), Int32(-1), dp_crit)
         # death is set well after the care ladder resolves
-        cc = CriticalHealthProfile(hospital_probability = 1.0, critical_onset_to_hospital_admission = 0,
-            hospital_to_icu_probability = 1.0, hospital_admission_to_icu_admission = 0,
+        cc = CriticalHealthProfile(hospital_probability = 1.0, severeness_onset_to_hospital_admission = 0,
+            hospital_to_icu_probability = 1.0, critical_onset_to_icu_admission = 0,
             icu_admission_to_icu_discharge = 5, icu_discharge_to_hospital_discharge = 3,
             death_probability = 1.0, critical_onset_to_death = 20)
         care2, outcome2 = calculate_health_profile(cc, ind, inf_crit, rng)
-        @test care2.hospital_admission == 8
-        @test care2.icu_admission == 8
+        @test care2.hospital_admission == 5    
+        @test care2.icu_admission == 8         
         @test care2.icu_discharge == 13
         @test care2.hospital_discharge == 16
         @test outcome2.death == 28
         @test care2.ventilation_admission == -1
 
         # full critical ladder: guaranteed ventilation nests inside ICU which nests inside the ward.
         # discharges chain inward-out, so ventilation ends first, then ICU, then hospital.
-        cc_vent = CriticalHealthProfile(hospital_probability = 1.0, critical_onset_to_hospital_admission = 0,
-            hospital_to_icu_probability = 1.0, hospital_admission_to_icu_admission = 0,
+        cc_vent = CriticalHealthProfile(hospital_probability = 1.0, severeness_onset_to_hospital_admission = 0,
+            hospital_to_icu_probability = 1.0, critical_onset_to_icu_admission = 0,
             icu_to_ventilation_probability = 1.0, icu_admission_to_ventilation_admission = 0,
             ventilation_admission_to_ventilation_discharge = 4, ventilation_discharge_to_icu_discharge = 2,
             icu_discharge_to_hospital_discharge = 3)
@@ -154,7 +154,7 @@ end
 
         # cascading-off caveat: hospital_to_icu_probability set without hospital_probability is a no-op,
         # since ICU is gated behind a hospital admission that never happens
-        cc2 = CriticalHealthProfile(hospital_to_icu_probability = 1.0, hospital_admission_to_icu_admission = 0,
+        cc2 = CriticalHealthProfile(hospital_to_icu_probability = 1.0, critical_onset_to_icu_admission = 0,
             icu_admission_to_icu_discharge = 5)
         care3, _ = calculate_health_profile(cc2, ind, inf_crit, rng)
         @test care3.hospital_admission == -1
@@ -232,7 +232,7 @@ end
     @testset "DefaultHealthProgression folds across a host's infections" begin
         rng = Xoshiro(1)
         cc = CriticalHealthProfile(hospital_probability = 1.0, hospital_to_icu_probability = 1.0, death_probability = 1.0,
-            critical_onset_to_hospital_admission = 0, hospital_admission_to_icu_admission = 0,
+            severeness_onset_to_hospital_admission = 0, critical_onset_to_icu_admission = 0,
             icu_admission_to_icu_discharge = 5, icu_discharge_to_hospital_discharge = 3,
             critical_onset_to_death = 9)
         hp = DefaultHealthProgression()
@@ -256,8 +256,11 @@ end
         filed = _filed(sched, 1)
         # two stays: hospital+ICU in, ICU+hospital out, twice over
         @test length(filed) == 8
-        @test (Int16(4), CARE_HOSPITAL, true) in filed     # pathogen 1, critical_onset 4
-        @test (Int16(20), CARE_HOSPITAL, true) in filed    # pathogen 2, critical_onset 20
+        # the ward is entered at severeness onset, the ICU at critical onset
+        @test (Int16(3), CARE_HOSPITAL, true) in filed     # pathogen 1, severeness_onset 3
+        @test (Int16(4), CARE_ICU, true) in filed          # pathogen 1, critical_onset 4
+        @test (Int16(6), CARE_HOSPITAL, true) in filed     # pathogen 2, severeness_onset 6
+        @test (Int16(20), CARE_ICU, true) in filed         # pathogen 2, critical_onset 20
         @test count(f -> f[2] === CARE_HOSPITAL && f[3], filed) == 2
 
         # earliest death wins
@@ -402,7 +405,7 @@ end
         hp = DefaultHealthProgression()
         # only pathogen 1 carries a profile; pathogen 2 misses the index entirely
         index = _idx((1, 0) => CriticalHealthProfile(hospital_probability = 0.4,
-            critical_onset_to_hospital_admission = 1, hospital_admission_to_hospital_discharge = 3))
+            severeness_onset_to_hospital_admission = 1, hospital_admission_to_hospital_discharge = 3))
 
         function stays(n, coinfect::Bool)
             rng = Xoshiro(1234); episodes = 0
@@ -436,9 +439,9 @@ end
             critical_onset_to_critical_offset = Poisson(5), critical_offset_to_severeness_offset = Poisson(3),
             severeness_offset_to_recovery = Poisson(10))
         crit = Critical(; dkw..., health = CriticalHealthProfile(
-            hospital_probability = 0.4, critical_onset_to_hospital_admission = Poisson(1),
+            hospital_probability = 0.4, severeness_onset_to_hospital_admission = Poisson(1),
             hospital_admission_to_hospital_discharge = Poisson(8),
-            hospital_to_icu_probability = 0.3, hospital_admission_to_icu_admission = Poisson(1),
+            hospital_to_icu_probability = 0.3, critical_onset_to_icu_admission = Poisson(1),
             icu_admission_to_icu_discharge = Poisson(5), icu_discharge_to_hospital_discharge = Poisson(2),
             death_probability = 0.15, critical_onset_to_death = Poisson(6)))
         pA = Pathogen(id = 1, name = "A", progressions = [crit],
@@ -713,9 +716,9 @@ end
                 "hospital_admission_to_hospital_discharge" => Dict("distribution" => "Poisson", "parameters" => [10])),
             "critical" => Dict(
                 "hospital_probability" => 0.9, "hospital_to_icu_probability" => 0.6, "death_probability" => 0.3,
-                "critical_onset_to_hospital_admission" => Dict("distribution" => "Poisson", "parameters" => [1]),
+                "severeness_onset_to_hospital_admission" => Dict("distribution" => "Poisson", "parameters" => [1]),
                 "hospital_admission_to_hospital_discharge" => Dict("distribution" => "Poisson", "parameters" => [10]),
-                "hospital_admission_to_icu_admission" => Dict("distribution" => "Poisson", "parameters" => [1]),
+                "critical_onset_to_icu_admission" => Dict("distribution" => "Poisson", "parameters" => [1]),
                 "icu_admission_to_icu_discharge" => Dict("distribution" => "Poisson", "parameters" => [8]),
                 "icu_discharge_to_hospital_discharge" => Dict("distribution" => "Poisson", "parameters" => [5]),
                 "critical_onset_to_death" => Dict("distribution" => "Poisson", "parameters" => [7])))

test/postprocessortest.jl

@@ -156,7 +156,7 @@ import GEMS: _mean_contacts_per_age_group,
             critical_onset_to_critical_offset = Poisson(5), critical_offset_to_severeness_offset = Poisson(3),
             severeness_offset_to_recovery = Poisson(10))
         crit = Critical(; dkw..., health = CriticalHealthProfile(
-            hospital_probability = 0.4, critical_onset_to_hospital_admission = Poisson(1),
+            hospital_probability = 0.4, severeness_onset_to_hospital_admission = Poisson(1),
             hospital_admission_to_hospital_discharge = Poisson(8),
             death_probability = 0.15, critical_onset_to_death = Poisson(6)))
         mkp(i, nm) = Pathogen(id = i, name = nm, progressions = [crit],

test/testdata/TestConf.toml

@@ -97,8 +97,8 @@
                 hospital_probability = 1.0
                 hospital_to_icu_probability = 1.0
                 death_probability = 0.3
-                critical_onset_to_hospital_admission = 0
-                hospital_admission_to_icu_admission = 0
+                severeness_onset_to_hospital_admission = 0
+                critical_onset_to_icu_admission = 0
 
                 # time to infectiousness (after exposure) [+1 will be added internally to avoid zero-day durations]
                 [Pathogens.Covid19.progressions.Critical.exposure_to_infectiousness_onset]