def assign_cv_outcome(self,
person,
years=1,
manualStrokeMIProbability=None):
return Outcome(
OutcomeType.STROKE,
False) if np.random.random() < self._stroke_rate else None
def assign_cv_outcome(self,
person,
years=1,
manualStrokeMIProbability=None):
return (Outcome(OutcomeType.MI,
np.random.random() < self._fatality_rate)
if np.random.random() < self._mi_rate else None)
def setUp(self):
self.joe = Person(
42,
NHANESGender.MALE,
NHANESRaceEthnicity.NON_HISPANIC_BLACK,
140,
90,
5.5,
50,
200,
25,
90,
150,
45,
0,
Education.COLLEGEGRADUATE,
SmokingStatus.NEVER,
AlcoholCategory.NONE,
0,
0,
0,
0,
initializeAFib,
)
self.joe_with_mi = copy.deepcopy(self.joe)
self.joe_with_mi.add_outcome_event(Outcome(OutcomeType.MI, False))
self.joe_with_stroke = copy.deepcopy(self.joe)
self.joe_with_stroke.add_outcome_event(Outcome(OutcomeType.STROKE, False))
self._population_dataframe = init_vectorized_population_dataframe(
[self.joe, self.joe_with_mi, self.joe_with_stroke],
with_base_gcp=True,
)
self._always_positive_repository = AlwaysPositiveOutcomeRepository()
self._always_negative_repository = AlwaysNegativeOutcomeRepository()
self.cvDeterminer = CVOutcomeDetermination(self._always_positive_repository)
def create_or_rollback_events_to_correct_calibration(
self, treatment_outcome_standard, treatedRisks, untreatedRisks,
outcomeType, fatalityDetermination, recalibration_pop):
modelEstimatedRR = treatedRisks.mean() / untreatedRisks.mean()
# use the delta between that effect and the calibration standard to recalibrate the pop.
delta = modelEstimatedRR - treatment_outcome_standard[outcomeType]
eventsForPeople = [
person.has_outcome_during_wave(self._currentWave, outcomeType)
for _, person in recalibration_pop.iteritems()
]
numberOfEventStatusesToChange = abs(
int(
round(delta * pd.Series(eventsForPeople).sum() /
modelEstimatedRR)))
nonEventsForPeople = [not item for item in eventsForPeople]
# key assumption: "treatment" is applied to a population as opposed to individuals within a population
# analyses can be setup either way...build two populations and then set different treatments
# or build a ur-population adn then set different treamtents within them
# this is, i thikn, the first time where a coding decision is tied to one of those structure.
# it would not, i think, be hard to change. but, just spelling it out here.
# if negative, the model estimated too few events, if positive, too mnany
if delta < 0:
if numberOfEventStatusesToChange > 0:
new_events = recalibration_pop.loc[nonEventsForPeople].sample(
n=numberOfEventStatusesToChange,
replace=False,
weights=pd.Series(
untreatedRisks).loc[nonEventsForPeople].values)
for i, event in new_events.iteritems():
event.add_outcome_event(
Outcome(outcomeType, fatalityDetermination(event)))
# redtag - two problems here...1. rolling back events in people that may not have events
# 2. probably usign the wrong weights...need to roll back inversely proportionately to the likeliood of an event, riht?
elif delta > 0:
if numberOfEventStatusesToChange > pd.Series(
eventsForPeople).sum():
numberOfEventStatusesToChange = pd.Series(
eventsForPeople).sum()
if numberOfEventStatusesToChange > 0:
events_to_rollback = recalibration_pop.loc[
eventsForPeople].sample(
n=numberOfEventStatusesToChange,
replace=False,
weights=pd.Series(
1 - untreatedRisks).loc[eventsForPeople].values)
for i, event in events_to_rollback.iteritems():
event.rollback_most_recent_event(outcomeType)
def test_fatal_mi_secondary_prob(self):
self.cvDeterminer.mi_secondary_case_fatality = 1.0
self.cvDeterminer.mi_case_fatality = 0.0
joeClone = copy.deepcopy(self.joe)
self.assertEqual(self.cvDeterminer._will_have_fatal_mi(joeClone, 0.0),
0)
joeClone._outcomes[OutcomeType.MI] = (joeClone._age,
Outcome(OutcomeType.MI, False))
# even though the passed fatality rate is zero, it shoudl be overriden by the
# secondary rate given that joeclone had a prior MI
self.assertEqual(self.cvDeterminer._will_have_fatal_mi(joeClone, 0.0),
1)
def assign_outcome_for_person(self,
outcome_model_repository,
person,
years=1,
manualStrokeMIProbability=None):
cvRisk = outcome_model_repository.get_risk_for_person(
person, OutcomeModelType.CARDIOVASCULAR, years=1)
if person._stroke or person._mi:
cvRisk = cvRisk * self.secondary_prevention_multiplier
if self._will_have_cvd_event(cvRisk):
if self._will_have_mi(person, outcome_model_repository,
manualStrokeMIProbability):
if self._will_have_fatal_mi(person):
return Outcome(OutcomeType.MI, True)
else:
return Outcome(OutcomeType.MI, False)
else:
if self._will_have_fatal_stroke(person):
return Outcome(OutcomeType.STROKE, True)
else:
return Outcome(OutcomeType.STROKE, False)
def get_outcome(self, person, mi, fatal, vectorized):
if vectorized:
person.miNext = mi
person.strokeNext = not mi
person.deadNext = fatal
person.miFatal = mi and fatal
person.strokeFatal = not mi and fatal
person.ageAtFirstMI = (
person.age
if (person.ageAtFirstMI is None) or (np.isnan(person.ageAtFirstMI))
else person.ageAtFirstMI
)
person.ageAtFirstStroke = (
person.age
if (person.ageAtFirstStroke is None) or (np.isnan(person.ageAtFirstStroke))
else person.ageAtFirstStroke
)
return person
else:
return Outcome(OutcomeType.MI if mi else OutcomeType.STROKE, fatal)
def update_person(self, person, x):
if person.is_dead():
raise Exception(f"Trying to update a dead person: {person}")
for rf in self._riskFactors:
attr = getattr(person, "_" + rf)
attr.append(x[rf + str(self._currentWave)])
for treatment in self._treatments:
attr = getattr(person, "_" + treatment)
attr.append(x[treatment + str(self._currentWave)])
# advance outcomes - this will add CV eath
for outcomeName, outcomeType in {
"stroke": OutcomeType.STROKE,
"mi": OutcomeType.MI,
"dementia": OutcomeType.DEMENTIA,
}.items():
if x[outcomeName + "Next"]:
fatal = False if outcomeName == "dementia" else x[outcomeName +
"Fatal"]
# only one dementia event per person
if outcomeName == "dementia" and person._dementia:
break
else:
person.add_outcome_event(Outcome(outcomeType, fatal))
person._gcp.append(x.gcp)
person._qalys.append(x.qalyNext)
person._bpMedsAdded.append(x.bpMedsAddedNext)
# add non CV death to person objects
if x.nonCVDeathNext:
person._alive.append(False)
# only advance age in survivors
if not x.deadNext:
person._age.append(person._age[-1] + 1)
person._alive.append(True)
return person
def get_dementia(self, person):
return Outcome(OutcomeType.DEMENTIA, False)
def assign_cv_outcome(self, person, years=1, manualStrokeMIProbability=None):
return Outcome(OutcomeType.MI, 0)
def __init__(
self,
age: int,
gender: NHANESGender,
raceEthnicity: NHANESRaceEthnicity,
sbp: int,
dbp: int,
a1c: float,
hdl: int,
totChol: int,
bmi: float,
ldl: int,
trig: int,
waist: int, # Waist circumference in cm
anyPhysicalActivity: int,
education: Education,
smokingStatus: SmokingStatus,
alcohol: AlcoholCategory,
antiHypertensiveCount: int,
statin: int,
otherLipidLoweringMedicationCount: int,
creatinine: float,
initializeAfib: Callable,
initializationRepository=None,
selfReportStrokeAge=None,
selfReportMIAge=None,
randomEffects=None,
**kwargs,
) -> None:
# building in manual bounds on extreme values
self._lowerBounds = {"sbp": 60, "dbp": 20}
self._upperBounds = {"sbp": 300, "dbp": 180}
self._gender = gender
self._raceEthnicity = raceEthnicity
self._alive = [True]
self._age = [age]
self._sbp = [self.apply_bounds("sbp", sbp)]
self._dbp = [self.apply_bounds("dbp", dbp)]
self._a1c = [a1c]
self._hdl = [hdl]
self._ldl = [ldl]
self._trig = [trig]
self._totChol = [totChol]
self._bmi = [bmi]
self._waist = [waist]
self._anyPhysicalActivity = [anyPhysicalActivity]
self._alcoholPerWeek = [alcohol]
self._education = education
# TODO : change smoking status into a factor that changes over time
self._smokingStatus = smokingStatus
self._antiHypertensiveCount = [antiHypertensiveCount]
self._statin = [statin]
self._otherLipidLoweringMedicationCount = [
otherLipidLoweringMedicationCount
]
self._creatinine = [creatinine]
# outcomes is a dictionary of arrays. each element in the dictionary represents
# a differnet outcome type each element in the array is a tuple representting
# the age of the patient at the time of an event (element zero). and the outcome
# (element one).multiple events can be accounted for by having multiple
# elements in the array.
self._outcomes = {
OutcomeType.MI: [],
OutcomeType.STROKE: [],
OutcomeType.DEMENTIA: []
}
self._selfReportStrokePriorToSim = 0
self._selfReportMIPriorToSim = 0
# a variable to track changes in BP meds compared to the baseline
self._bpMedsAdded = [0]
# convert events for events prior to simulation
if selfReportStrokeAge is not None and selfReportStrokeAge > 1:
self._selfReportStrokeAge = (selfReportStrokeAge if
selfReportStrokeAge <= self._age[-1]
else self._age[-1])
self._selfReportStrokePriorToSim = 1
self._outcomes[OutcomeType.STROKE].append(
(-1, Outcome(OutcomeType.STROKE, False)))
if selfReportMIAge is not None and selfReportMIAge > 1:
self._selfReportMIAge = (selfReportMIAge
if selfReportMIAge <= self._age[-1] else
self._age[-1])
self._selfReportMIPriorToSim = 1
self._outcomes[OutcomeType.MI].append(
(-1, Outcome(OutcomeType.MI, False)))
for k, v in kwargs.items():
setattr(self, k, v)
if initializeAfib is not None:
self._afib = [initializeAfib(self)]
else:
self._afib = [False]
# for outcome mocels that require random effects, store in this dictionary
self._randomEffects = {"gcp": 0}
if randomEffects is not None:
self._randomEffects.update(randomEffects)
# lucianatag: for this and GCP, this approach is a bit inelegant. the idea is to have classees that can be swapped out
# at the population level to change the behavior about how people change over time.
# but, when we instantiate a person, we don't want to keep a refernce tot the population.
# is the fix just to have the population create people (such that the repository/strategy/model classes can be assigned from within
# the population)
self._qalys = []
self._gcp = []
if initializationRepository is not None:
initializers = initializationRepository.get_initializers()
for initializerName, method in initializers.items():
attr = getattr(self, initializerName)
attr.append(method(self))
self._bpTreatmentStrategy = None
def get_dementia(self, person):
if npRand.uniform(size=1) < self.get_risk_for_person(
person, OutcomeModelType.DEMENTIA):
return Outcome(OutcomeType.DEMENTIA, False)
else:
return None
def __init__(
self,
age: int,
gender: NHANESGender,
raceEthnicity: NHANESRaceEthnicity,
sbp: int,
dbp: int,
a1c: float,
hdl: int,
totChol: int,
bmi: float,
ldl: int,
trig: int,
waist: int, # Waist circumference in cm
anyPhysicalActivity: int,
education: Education,
smokingStatus: SmokingStatus,
alcohol: AlcoholCategory,
antiHypertensiveCount: int,
statin: int,
otherLipidLoweringMedicationCount: int,
initializeAfib: Callable,
selfReportStrokeAge=None,
selfReportMIAge=None,
**kwargs,
) -> None:
# building in manual bounds on extreme values
self._lowerBounds = {"sbp": 60, "dbp": 20}
self._upperBounds = {"sbp": 300, "dbp": 180}
self._gender = gender
self._raceEthnicity = raceEthnicity
self._alive = [True]
self._age = [age]
self._sbp = [self.apply_bounds("sbp", sbp)]
self._dbp = [self.apply_bounds("dbp", dbp)]
self._a1c = [a1c]
self._hdl = [hdl]
self._ldl = [ldl]
self._trig = [trig]
self._totChol = [totChol]
self._bmi = [bmi]
self._waist = [waist]
self._anyPhysicalActivity = [anyPhysicalActivity]
self._alcoholPerWeek = [alcohol]
self._education = education
# TODO : change smoking status into a factor that changes over time
self._smokingStatus = smokingStatus
self._antiHypertensiveCount = [antiHypertensiveCount]
self._statin = [statin]
self._otherLipidLoweringMedicationCount = [
otherLipidLoweringMedicationCount
]
# outcomes is a dictionary of arrays. each element in the dictionary represents
# a differnet outcome type each element in the array is a tuple representting
# the age of the patient at the time of an event (element zero). and the outcome
# (element one).multiple events can be accounted for by having multiple
# elements in the array.
self._outcomes = {OutcomeType.MI: [], OutcomeType.STROKE: []}
self._selfReportStrokePriorToSim = 0
self._selfReportMIPriorToSim = 0
# convert events for events prior to simulation
if selfReportStrokeAge is not None and selfReportStrokeAge > 1:
self._selfReportStrokePriorToSim = 1
self._outcomes[OutcomeType.STROKE].append(
(-1, Outcome(OutcomeType.STROKE, False)))
if selfReportMIAge is not None and selfReportMIAge > 1:
self._selfReportMIPriorToSim = 1
self._outcomes[OutcomeType.MI].append(
(-1, Outcome(OutcomeType.MI, False)))
for k, v in kwargs.items():
setattr(self, k, v)
if initializeAfib is not None:
self._afib = [initializeAfib(self)]
else:
self._afib = [False]
self._gcp = []
# for outcome mocels that require random effects, store in this dictionary
self._randomEffects = dict()
self._bpTreatmentStrategy = None
def assign_cv_outcome(self,
person,
years=1,
manualStrokeMIProbability=None):
return Outcome(OutcomeType.STROKE, 1) if person._age[-1] > 50 else None
def assign_cv_outcome(self, person):
return Outcome(OutcomeType.MI, False)
