A Study of Concurrency Robert Puckett UH Manoa November 20 2014 Outline Core Concepts HIV Concurrency MSM Agents MASHIV System Design HIV Model Agents Sexual Negotiation Modes of ID: 799379
Download The PPT/PDF document "MASHIV Multi-Agent Simulation of HIV in ..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
MASHIV
Multi-Agent Simulation of HIV in MSM CommunitiesA Study of Concurrency
Robert Puckett, UH
Manoa
, November 20, 2014
Slide2Outline
Core ConceptsHIVConcurrencyMSMAgents
MASHIV
System Design
HIV ModelAgentsSexual NegotiationModes of OperationQ & A
2
Slide3Research Questions
Can a multi-agent simulation of individual level HIV transmission illuminate the impact that concurrent sexual relationships have on the HIV epidemic in MSM communities?What is the impact of PrEP amid concurrency
on the
resulting HIV epidemic?
Can we overcome stochastic variability to provide consistent analysis and recommendations?3
Slide4Core concepts
Really Quick Overview
Slide5HIV Stages
Primary/Acute HIV Infection (PHI)Virulence: HighEstimated 43% of infections due to PHI period [Wawer]
Occurs 2-4 weeks after exposure
Duration: 1.5 – 12 months
[Blaser, 2013]Only 2/3 experience symptomsFever, fatigue, pharyngitis, weight loss, night sweats, lymphadenopathy, myalgia, headache, nausea
Symptoms commonly result in misdiagnosis
Correctly diagnosed as PHI in 1000 of 60 million cases
HIV undetectable by antibody tests for several weeks into PHI
[
Coplan
]
5
Slide6HIV Stages
Asymptomatic PeriodVirulence: Low, but presentDuration: 8+ years, untreatedAcquired Immuno-deficiency Syndrome (AIDS)
Immune system badly damaged (CD4 count < 200)
Susceptible to
Opportunistic infectionsCertain cancersLife expectancy1-3 years depending on presence of opportunistic infections
6
Slide7Viral Plasma Load by Stage
7
Slide8Concurrency
”Overlapping sexual partnerships where sexual intercourse with one partner occurs between two acts of intercourse with another partner.” [UNAIDS, 2009]Key ComponentsDuration of relationships
Contact frequency
Number of partners
Virulence of partners
Serial Monogamy
Concurrent partnerships
1
2
3
4
5
time
1
2
3
4
5
time
8
Slide9Concurrency & PHI
The concernConcurrent relationships will create a highly connected/reachable sexual networkPHI stage is highly virulent, unlikely to be detectedResult
Waves of PHI stage HIV infection sweep through sexual network
concurrency
monogamy
9
Slide10The MSM Community
MSM – Behavior specific termIncludes gay, bisexual, male sex workers, transgenderedAny other men who engage in same-gendered sexStatistics
51
% of new HIV cases in the US were MSM
14-19% of urban MSM are HIV+ [Goodreau 2007]HIV prevalence in MSM increasing in most developed countries since 90s [Grulich 2008]
10
Slide11MSM & HIV Risk
Risk FactorsUnprotected Anal Receptive Sex (UAR)Disproportionately affected by STIs [Goodreau
]
Other behavioral factors
“Bareback “ sex seeking, sero-sorting, high-risk venuesHigher proportion of concurrent relationships?Concurrent relationshipsGeneral US Population: 11% of men
San Francisco urban MSM cohort: 78% of men
Note: vastly different sample populations
Less sexual role segregation
Heterosexual sexual role defined by gender
MSM sexual role defined by preference (
insertive
/receptive)
Role versatility allows HIV to spread more easily
11
Slide12Multi-agent systems
An Overview
Slide13What are Agents?
A programming construct meant to represent a real-life entity or roleKey characteristicsAutonomousGoal orientedSelf-organizing
Exist in / React to environment
Local knowledge
DecentralizedCompete, coordinate, cooperateIn an individual-level multi-agent simulation of HIV1 agent == 1 person
13
Slide14Multi-Agent System
The AgentsCognitive modelPersonal history of interactions with the world/agentsThe EnvironmentObservable by the agents
Does not directly control the agents
The Rules
Actions the agents may chooseReactions to the environment/agentsGoal-oriented behaviorLimitations on agent behavior
14
Slide15Why use Agents for HIV?
Intuitive pairing of agents and peopleSimple rules can result in complex behaviorAllows for observing the dynamics of individual-level decision making on the HIV epidemicPotentially useful for guiding real-world studies and interventions
15
Slide16MASHIV
The Multi-Agent Simulation for HIV Transmission
Slide17MASHIV
GoalUse JAVA to develop a multi-agent simulation of HIV for the MSM communityDetermine the role of concurrency in HIV epidemics of MSMTrack and AnalyzeHIV Prevalence/Incidence trends
Proportion of infections resulting from PHI
Concurrency measure of population over time
17
Slide18MASHIV Operation
User defines parameter setGlobal vs. Population parameter assignmentPopulation definition
Runtime
Initialization
Process User ParametersGenerate Relationship SchemasGenerate AgentsMain LoopUpdate AgentsUpdate HIV Disease ProgressionUpdating Existing Relationships
Sex, HIV Transmission, Relationship Ending
Date & Add Relationships
Statistics Collection
18
Slide19Parameters
19
Slide20Person Agent
Represents an MSM personForms/Ends RelationshipsEvaluates potential partnersReacts to HIV infection
20
Slide21Relationship Instance
Agents can have different schemas and expectations for relationship.21
Slide22Relationship Schema
Duration vs. Probabilistic Mode22
Slide23Relationships
Wanting to DateModesDuration-based: Time since last dateProbabilistic: Probability of formationFactors
Existing steady relationship
Number of relationships
Dating Pools10 random dating agentsSexual role compatibleBoth seeking same relationship type (Casual, Steady)Steady: no repeats; Casual: repeats allowedDating Evaluation…
23
Slide24Dating Evaluation
24
Slide25MASHIV hiv model
The Multi-Agent Simulation for HIV Transmission
Slide26HIV Model
InitializationDistributed to population based upon user inputHIVInstance is created
Transmission
Risk
Viral load of stage determines virulenceSafe sex practice determines riskProgression toward mortalityCD4 compartment model
26
Slide27HIV Instance
Handles HIV progression & mortalityLogs key informationSource of InfectionStage of infection source
27
Slide28HIV Transmission
28Condom UseTransmission Risk
S – Stage risk multiplier
Art() - ART risk reduction
Slide29HIV Transmission
29
Reception Risk
P
() : Sexual role riskPrEP
() –
PrEP
risk reduction
Circ
() – Circumcision risk reduction
Slide30HIV Infection
Fast-moving vs. Regular SpeedVirus StagesPHI – 90 daysAsymptomatic & AIDS
CD4 Model
30
Slide31HIV Progression
CD4 Compartment ModelAdapted from Spectrum/EPP specification [UNAIDS]Used to progress agents from infection to death
Compartments correlate to CD4 counts of individuals
Annual progression rates define progression between compartments and compartment mortality
Reduces progression rate to account for ART usage
31
Slide32Lambdas
32
Slide33Mus
33
Slide34Alphas
34
Slide35Transmission Probabilities
Tool for viewing transmission probabilities for model35
Slide36MASHIV Interface
A Multi-Agent Simulation of HIV
Slide37Main Window & Menus
37
Slide38Parameter Set
38
Slide39Parameter Set
39
Slide40Interactive Dash
40
Slide41Interactive Dash - Running
41
Slide42Interactive Dash
42
Slide43Interactive Dash – Adding Set
43
Slide44Aggregate Set Editor
44
Slide45Aggregate Set - Running
45
Slide46Aggregate Set - Running
46
Slide47Progression in MASHIV
Analysis tool for observing Stage & Group Progression in Model47
Slide48Mortality In MASHIV
Tool for observing mortality without ART
48
Slide49Mortality In MASHIV
Tool for analyzing mortality with ARTGraphs represent ART started in first CD4 group assignedAssumes only mortality based ART failureLack of background mortality
49
Slide50Q&A
Questions and Answers