Ny kunskap f

1
PROSPER Smittsamma sjukdomars inverkan på
det svenska samhället mot evidensbaserade
responsstrategier

• Toomas Timpka Henrik Eriksson Elin A
  Gursky
• Anders Grimvall Joakim Ekberg Olle
  Eriksson
• Magnus Strömgren Einar Holm Lars
  Valter
• James M Nyce

2
The CriSim group

• Department of Computer and Information Science
• Department of Medicine and Health Sciences
• Linköping University, Linköping, Sweden
• ANSER/Analytic Services Inc.
• Arlington, VA., USA
• Department of Social and Economic Geography
• Umeå University, Umeå, Sweden
• Department of Anthropology
• Ball State University, Muncie, IN., USA

3
Introduction

• Over the last decades, several serious infectious
  diseases have emerged rapidly to become global
  threats, including the severe acute respiratory
  syndrome (SARS) and avian influenza.
• Each of these diseases has required a fast and
  specific response from policy-makers and public
  health authorities.
• Such a situation occurred in 2009 with the
  emergence of the of a novel A/H1N1 influenza
  virus (the swine flu) in Mexico.

4
Introduction the swine flu

• On 28 April, Mexico had 26 confirmed human cases
  with seven confirmed deaths. Elsewhere, there
  were confirmed cases and deaths in the USA,
  Canada, UK, Spain, New Zealand, and Israel (1).
• On 11 May, researchers analyzing data from the
  Mexican outbreak reported that the
  transmissibility of the new influenza strain was
  substantially higher than for seasonal flu, and
  comparable with lower estimates of the basic
  reproduction rate obtained from previous
  influenza pandemics (2).
• One month later, transmission in several
  countries could no longer be traced to
  clearly-defined chains of human-to-human
  contacts, and further spread was considered
  inevitable. Accordingly, with 30.000 confirmed
  cases in 74 counties, the WHO raised the
  influenza pandemic alert to the highest level,
  phase 6 (WHO 2009)
• 1. Swine influenza how much of a global threat?
  Lancet 2009373(9674)1495.
• 2. Fraser C, Donnelly CA, Cauchemez S, et al.
  Pandemic Potential of a Strain of Influenza A
  (H1N1) Early Findings. Science 2009. DOI
  10.1126/science.1176062

5
Pandemic influenza

• Global epidemic with high mortality
• Spanish influenza (1918)
• Asian flu (1957)
• Hongkong flu (1968)
• Modified (1918) or reassorted (1957, 1968) avian
  virus
• Destructive in non-traditional risk groups
• Taubenberger et al, Nature 2005437889-892.

Edward Munch After the Spanishdisease (self
portrait), 1919.
6
Introduction public health response

• Public health officials in nations affected
  by the swine flu decided on response actions
• In the absence of a vaccine, closure of schools
  with infected pupils was used by some countries,
  but not others
• In the USA, the CDC initially supported school
  closures
• The Public Health Agency of Canada did not
  recommend closing schools
• The UK Health Protection Agency took the position
  that consideration should be given to
  temporarily closing the school (1)
• 1. Editorial. Putting influenza A H1N1 in its
  place. Lancet Infectious Diseases.
  2009DOI10.1016/S1473-3099(09)70134-3 1.

7
Introduction infrastructural issues

• The principles for forming the response to the
  swine flu seem to have remained
  unsystematically linked to the particular forms
  of disease validation and prediction locally on
  hand.
• If poorly validated and coordinated methods are
  allowed to inform policy-making on emerging
  infectious diseases, these methods may
  dangerously mislead critical response
  implementation (1,2)
• The social distancing example from the swine
  flu outbreak is particularly disquieting in
  light of that the effectiveness of this set of
  measures recently had been questioned because of
  the major impediments to compliance (3).
• 1. ECDC. Now-casting and short-term forecasting
  during influenza pandemics - a focused
  developmental ECDC workshop. Stockholm ECDC,
  2007.
• 2. Timpka T, Eriksson H, Gursky E, et al.
  Population-based simulations of influenza
  pandemics validity and significance for public
  health policy. Bull World Health Organ
  200987305-311.
• 3. Rothstein MA, Talbott MK. Encouraging
  compliance with quarantine a proposal to provide
  job security and income replacement. Am J Public
  Health 200797 Suppl 1S49-56.

8
Overview of workflow in establishment of evidence
on rapidly emerging infectious disease outbreaks
9
Research problems

• There is no common infrastructure in place for
  analyses of pandemic data and sharing of
  information
• The co-ordination of the response within nations
  and across national borders remains an issue
• An information infrastructure is required
  that differs from traditional health information
  systems in that it is to be used in situations
  when infectious diseases overwhelm the
  first-order resources on hand designed to detect
  and control the outbreak.

10
Research aims

• to draft a protocol that can be used to realize a
  standardized information infrastructure for rapid
  production of pandemic response program evidence
• The hypothesis is that it is possible to
  standardize the establishment of a distributed
  global infrastructure for rapidly translating
  evidence from analyses of available data into
  coordinated response when addressing worldwide
  emerging infectious diseases.

11
Methods

• Overview of the methods used for
• A. Data collection
• B. Data analysis

12
Methods

• Data collection
• A nominal group method (1) was used to collect
  requirements data.
• Two expert panels examined and outlined
  requirements on the protocol with regard to the
  scope of data sources and analytic functions to
  be covered, respectively.
• Individual experts reviewed a working
  requirements document followed by telephone
  discussions (n18).
• Requirements on the data were defined by a panel
  consisting of scientists and practitioners (n8)
  with backgrounds in medicine, epidemiology,
  medical anthropology, computer science, health
  informatics, cognitive science, and
  socio-economic geography.
• The panel examining requirements on analytic
  functions consisted of scientists and
  practitioners (n5) with backgrounds in medicine,
  statistics, computer science, health informatics,
  and cognitive science.
• When subsequent turns did not return significant
  changes in the documents, the requirement
  specifications were considered to be established.
  
• 1, Jones J, Hunter D. Consensus methods for
  medical and health services research. Bmj
  1995311(7001)376-80.

13
Methods

• Data analysis
• A method for rational solution of multi-facetted
  design problems (1,2) was used for data analysis
• The members of the two panels were merged into
  one protocol specification group
• The task communicated to the group was to
  formulate a protocol design using the
  requirements, their subject matter expertise, and
  the published literature. The experts first
  provided their individual comments, which were
  collected by a design process coordinator
• Formulation of functional design solutions was
  performed independently by experts who reviewed a
  document describing the model described as design
  patterns
• The design patterns were represented in the form
  Title, Problem-Requirements, Functional design,
  and Realization
• In the final step, the design patterns were
  summarized in the PROSPER protocol (PROtocol for
  Standardized Pandemic and Emerging infectious
  disease Response)
• 1. Rittel H, Webber M. Dilemmas in a general
  theory of planning. Policy Sciences
  19734(2)55-169.
• 2. Simon H. Design of the artificial. 2nd ed.
  Cambridge, Mass. The MIT Press, 1981.

14
Results - requirements

• Status overview Major problems during the early
  stages of pandemic planning include
• - analytic methods and technologies are
  uncoordinated with the organization of response
  programs
• - shortage of reliable microbiological and
  epidemiological data, and
• - lack of universally applicable detection
  and forecasting methods
• Data requirements Specifications are needed of
  population and community data, the quality and
  timeliness of outbreak data, and on how data on
  population behavior are represented, especially
  over time
• Analytic functions requirements Comparative
  assessments of response program effectiveness,
  rather than efficacy, are needed

15
Results PROSPER

• The PROSPER protocol outlines an information
  infrastructure for pandemic response that is
  defined with reference to response program
  implementation.
• The infrastructure can be realized using
  conventional system implementation methods by
  regional, national and international public
  health agencies or other organizations with an
  interest in rapidly responding to pandemic
  threats.
• PROtocol for Standardized Pandemic and Emerging
  infectious disease Response

16
Results

• Display of the PROSPER protocol in relation to
  intervention program implementation (1)
• 1. Rogers E. Diffusion of Innovations. Fifth ed.
  New York Free Press, 2003.

17
Results

• The PROSPER infrastructure covers analyses of the
  response context, the response processes, and
  outcomes and impacts. For each of these aspects,
  it outlines the basic infrastructure at levels of
  evidence, function, and technical systems.
• Evidence on the program context is organized with
  reference to the STROBE, guidelines on what
  should be included in reports of observational
  studies
• Reports on the process design and program
  effectiveness are organized according to the
  SQUIRE guidelines for reporting studies of
  quality improvement in health services

18
(No Transcript)
19
Results

• The structure of the functional level reflects
  the methods used to produce pandemic evidence and
  the organization of infectious disease response.
• The Capacity and needs assessment function is
  informed at the technical systems level be
  systems for laboratory and syndromic data access
  and visualization, and analysis scenario
  management (Table 1).

20
(No Transcript)
21
Results

• Table 1. PROSPER design pattern Capacity and
  needs assessment.
• Capacity and needs assessment
• Analysis scenario management
• Problem-Requirements A1 Sociogeographical
  representation of communities
• Functional design A spatially explicit model for
  population representation is used to allow for
  experiments in full scale with factual and
  synthetic populations. This design solution
  supports that different sociogeographical
  scenarios can be defined by changing the starting
  conditions of the model. Other basic model
  categories included in the pandemic outbreak
  scenario define transportation systems and other
  geographic conditions, e.g. location of
  workplaces, schools, and facilities for sports
  and entertainment events. The representations
  also include relational variables, such as
  individual-mother, -partner, -child, and
  co-worker at workplace. These relational
  variables are directly relevant for
  representation of the social networks
  transmitting infectious agents. In other words,
  sociogeographical preprocessing of spatially
  explicit population data are used to in advance
  identify specific groups and populations that may
  require more careful and intensified
  surveillance. The Strengthening the Reporting of
  Observational Studies in Epidemiology (STROBE)
  recommendations (1) on what should be included in
  accurate reports of observational studies are
  used to organize the communication of evidence
  from the analyses.
• Realization The scenario management is based on
  the ontology handling system Protégé (2). In
  addition to the SVERIGE (System for Visualizing
  Economic and Regional Influences Governing the
  Environment) model for sociogeographical
  representation of the Swedish population (3),
  preliminary settings for additional scenario
  models have been developed, representing, e.g.
  local social interaction and commuting patterns
  (4).
• Access to epidemiological and population data
• Problem-Requirements A2 Control and
  visualization of data quality and timeliness
• Design solution Epidemiological data from factual
  outbreaks are complemented with artificially
  generated data and collected into databases for
  use in detailed analyses of historical outbreaks
  and experiments on hypothetical outbreaks in
  populations. The factual outbreak data range
  from highly specific genomic and microbiological
  laboratory data to non-specific syndromic data,
  e.g. from telephone health advice centers and
  Internet website logs (5, 6). All data sources
  are controlled by methods for systematic
  statistical follow-up of the data used. In
  particular short-term trends in the pandemic
  progress can easily be masked by errors in
  sampling or laboratory practices. Statistical
  tools for trend analysis, such as semiparametric
  regression models (7), are therefore used to
  identify causes to flaws in the data collection
  routines that can lead to erroneous
  interpretations.
• Realization Population-based administrative
  healthcare databases (8) are used to assemble
  geographically explicit data from infections
  disease outbreaks at local levels. In Sweden, the
  regional telephone health advice services have
  been synchronized into a national call center
  supported by a telehealth Electronic Patient
  Record (EPR), where the reason for contact and
  residence for each caller is documented using a
  controlled terminology. We use the national
  database collecting data from all regional
  telehealth EPRs as a source for syndromic
  surveillance data. Regarding visualization
  services, these have not been included in the
  present realization. Instead, interactive graphs
  (www.ggobi.org) and motion chart
  (www.gapminder.org) services available at the
  Internet are used for getting overviews of large
  data sets.
• References
• 1. von Elm E, Altman DG, Egger M, Pocock SJ,
  Gøtzsche PC, Vandenbroucke JP STROBE Initiative.
  The Strengthening the Reporting of Observational
  Studies in Epidemiology (STROBE) statement
  guidelines for reporting observational studies. J
  Clin Epidemiol. 2008 Apr61(4)344-9.
• 2. Gennari JH, Musen MA, Fergerson RW, Grosso WE,
  Crubézy M, Eriksson H, et al. The evolution of
  Protégé An environment for knowledge-based
  systems development. Int J Hum Comp Stud
  200358(1)89-123.
• 3. Holm E, Holme K, Mäkilä K, Mattsson-Kauppi M,
  Mörtvik G. The SVERIGE Spatial Microsimulation
  Model Content, Validation, and Example
  Applications. GERUM 20024.
• 4. Holm E, Timpka T. A discrete time-space
  geography for epidemiology from mixing groups to
  pockets of local order in pandemic simulations.
  Stud Health Technol Inform 2007129464-8.
• 5. Sintchenko V, Gallego B. Laboratory-guided
  detection of disease outbreaksthree generations
  of surveillance systems. Arch Pathol Lab Med.
  2009 Jun133(6)916-25.
• 6. Smith GJ, Vijaykrishna D, Bahl J, Lycett SJ,
  Worobey M, Pybus OG, Ma SK, Cheung CL, Raghwani
  J, Bhatt S, Peiris JS, Guan Y, Rambaut A. Origins
  and evolutionary genomics of the 2009
  swine-origin H1N1 influenza A epidemic. Nature.
  2009 Jun 11. Epub ahead of print PubMed PMID
  19516283.
• 6. Wahlin K, Grimvall A. Uncertainty in water
  quality data and its implications for trend
  detection lessons from Swedish environmental
  data. Environmental Science and Policy
  200811115-124.
• 7. Wirehn AB, Karlsson HM, Carstensen JM.
  Estimating disease prevalence using a
  population-based administrative healthcare
  database. Scand J Public Health 200735(4)424-31.

22
Results

• The functions for Implementation process
  evaluation are supported by technical systems for
  response process analysis and knowledge-base
  maintenance (Table 2).

23
(No Transcript)
24
Results

• Table 2. PROSPER design pattern Implementation
  process evaluation (section)
• Implementation process evaluation
• Iterative response program implementation
• Problem-Requirements A3 Explicit representation
  of populations over time, D1 Adjustments to
  missing data, I2 Explicit fact and hypothesis
  management
• Functional design An iterative procedure for
  response program design was envisioned, where
  analyses of virtual outbreak detection and
  simulated interventions are used until real-time
  surveillance data and evaluations of factual
  interventions become available. The disease
  models used in the virtual analyses are
  preliminarily instantiated from the literature,
  e.g. with regard to incubation period and serial
  interval. Thereafter, program components are
  specified in intervention models. Response
  program developers can prepare process analyses
  by configuring program components and specifying
  intervention model parameters, e.g. the
  prophylactic performance of specific antiviral
  drugs or combinations. The SQUIRE guidelines for
  reporting studies of quality improvement in
  health services (1) are used for communication of
  evidence from the analyses.
• Realization The engineering principle underlying
  the example system is separation between the
  software for management of the outbreak models
  and the software for execution of the analyses
  (2). This separation allows flexible modeling to
  represent unexpected events and circumstances,
  while maintaining the run-time performance of
  outbreak detection and simulation programs. Basic
  disease and intervention characteristics are
  available from profiles reported in the
  literature (3) and at the Internet
  (https//www.epimodels.org/midas/modelProfilesFull
  .do). The basic models are combined and for a
  typology of explicit models and baseline
  parameter settings.
• References
• 1. Davidoff F, Batalden P, Stevens D, Ogrinc G,
  Mooney S SQUIRE development group. Publication
  guidelines for quality improvement in health
  care evolution of the SQUIRE project. Qual Saf
  Health Care. 2008 Oct17 Suppl 1i3-9.
• 2. Eriksson H, Morin M, Jenvald J, Gursky E, Holm
  E, Timpka T. Ontology based modeling of pandemic
  simulation scenarios. Stud Health Technol Inform
  2007129755-9.
• 3. Carrat F, Vergu E, Ferguson NM, Lemaitre M,
  Cauchemez S, Leach S, et al. Time lines of
  infection and disease in human influenza a
  review of volunteer challenge studies. Am J
  Epidemiol 2008167(7)775-85.

25
Results

• The last section of the protocol, Outcome and
  impact evaluation outlines the details for the
  comparative analyses of outbreak algorithms and
  the assessment of intervention effectiveness
  (Table 3).

26
(No Transcript)
27
Results

• The functions for comparative analyses in the
  outcome and impact evaluation section are based
  on technical systems for simulations and
  statistical analyses that can be acquired without
  major financial investments, by utilizing
  knowledge based systems techniques and networked
  cloud computing

28
PROSPER examples A two-tier model of epidemic
progression the biological tier
Incubation 1-3 days (average 1.9 days)
Contagious 3-6 days (average 4.1 days)
29
PROSPER examples A two-tier model of epidemic
progression the sociogeographical tier

• Geographical
• Logistical
• Social
• Cultural

• Mixing network approach to represent meeting
  places as social pockets
• Households are central social pockets

30
Day rest with symptoms Decision
modifiers Personal care resources Healthcare
access Health beliefs
Symptoms development
Non-intervenable conditions modifying pandemic
spread Individual person Sex Age Genetic
constitution Ethnicity Formal education Employment
Physical environment Climate Urbanisation
level Transportation network Social
environment Community Market and economy Social
capital Family Financial resources Family
structure and roles Social network
Intervenable mechanisms mediating pandemic spread
Individual person Prevention knowledge Compliance
to policies Self-protective behaviour Immunizatio
n Nutritional status Physical environment Healthc
are facilities Schools Workplaces Sanitary
standard Information infrastructure Social
environment Community Laws and regulations
Pandemic plans Mass media Family Family
behaviour
Individual person Regular social interaction in
personal social pockets in the
community Exposure to infectious individuals
Not infected
Asymptomatic infection

• Day rest
• Decision modifiers
• Health beliefs
• Mental models
• of pandemic
• of society

Social geographic data and assumptions regarding
social order
31
PROSPER examples Management of two-tier
simulation models - the PROTEGÉ ontology
handling system
32
PROSPER examples Cloud computing for outsourcing
of complex computations

• Simulations are defined in the ontology
  management system and thereafter distributed to
  anetworked computing environment

33
We set out to compute a preliminary Neighborhood
Influenza Susceptibility Index (NISI) describing
the vulnerability of local communities of
different geo-socio-physical structure to a
pandemic influenza outbreak.

• Responding to the current pandemic and preparing
  for future ones requires critical planning for
  the early phases where there is no availability
  of pandemic vaccine.

34
The Neighborhood Influenza Susceptibility Index
(NISI)

• The aim was to pre-compute maps describing local
  variations between geographical areas with regard
  to susceptibility to influenza transmission.
• The maps can be employed by local public health
  officials for planning of response measures
  before factual transmission data are at hand.
• Specifically, computation of a preliminary
  Neighborhood Influenza Susceptibility Index
  (NISI) is used to describe the vulnerability of
  local communities of different geo-socio-physical
  structure to a pandemic influenza outbreak.
• In difference to seasonal influenza, the herd
  immunity to a pandemic is by definition low,
  leading to that disease transmission largely is
  determined by the pattern of social contacts in
  the community.

• The NISI is estimated from a standardized virtual
  outbreak. One person per 1000 individuals in the
  fully susceptible study community is randomly
  selected and infected at t0, defined to be 9am
  the first day of the simulation.
• The rates of secondary infected individuals in
  different neighborhoods and the sociodemographic
  characteristics of these cases are thereafter
  recorded during the progress of the outbreak.
• Neither behavioral changes nor any further
  introduction of infected individuals by commuting
  and national or international travel are expected
  to take place during the standardized virtual
  outbreak.
• The preliminary NISI is finally computed as the
  proportion of infected at the end of the virtual
  outbreak.

35

• Geographical distribution of neighborhoods in the
  study municipality

36

• Epidemic curves for the virtual outbreaks (n10)
  generated for Linköping municipality

37

• Epidemic curves for the standardized outbreak
  displayed by neighborhood (n13).

38

• Numbers of uninfected individuals at selected
  days of the virtual outbreak displayed by
  neighborhood

39

• Aggregate-level neighborhood socioeconomic data
  in percent displayed by descending NISI/H1N1 t120
  (proportion of infected individuals) at the end
  of the virtual outbreak.

40
Discussion

• We have drafted the PROSPER protocol that can be
  used to realize a standardized information
  infrastructure for rapid production of pandemic
  response program evidence in different
  organizational and technical settings.
• The protocol is optimized with regard to analyses
  of response program effectiveness in particular
  communities and populations worldwide.
• In areas such as urban planning, pattern
  languages have been extensively used to transfer
  value-bearing design features between different
  milieus (1)
• 1. Alexander C. The timeless way of building.
  Oxford Oxford University Press, 1979.

41
Discussion

• It is necessary to caution against
  over-interpretation of predictive modeling
  results in policy-making settings, even in
  situations where different models display similar
  effectiveness of interventions (1).
• Analyses based on explorative modeling tools of
  what-if analysis type, such as FluAid and
  FluSurge/ have previously been used to directly
  inform policy recommendations concerning hospital
  surge capacity (2) and loss of medical work time
  (3) when planning pandemic responses.
• These modeling environments are not adjusted to
  the requirement that health intervention programs
  must be evidence-based.
• The PROSPER protocol is specifically adapted to
  that without being able to inspect, understand
  and adjust baseline assumptions, it is not clear
  to what extent policy makers will use, let alone
  trust and rely on, the analytic resources
  included in the infrastructure.
• 1. Halloran ME, Ferguson NM, Eubank S, Longini
  IM, Jr., Cummings DA, Lewis B, et al. Modeling
  targeted layered containment of an influenza
  pandemic in the United States. Proc Natl Acad Sci
  U S A 2008.
• 2. Ten Eyck RP. Ability of regional hospitals to
  meet projected avian flu pandemic surge capacity
  requirements. Prehosp Disaster Med
  200823(2)103-12.
• 3. Wilson N, Baker M, Crampton P, Mansoor O. The
  potential impact of the next influenza pandemic
  on a national primary care medical workforce. Hum
  Resour Health 200537.

42
Discussion limitations

• As for the technical systems level of PROSPER,
  some items identified in the requirements
  analysis were not covered by the present version
  of PROSPER.
• A technology that can support reliable,
  short-term forecasts is nowcasting, i.e.
  short-term predictions that rely on
  straight-forward extrapolation of recent
  observations in time (1).
• Early identification of the virus genome is
  central in the response to pandemic influenza
  (2). The laboratory system infrastructure is not
  in detail included in the present version of
  PROSPER.
• 1. Wilson N, Baker M, Crampton P, Mansoor O. The
  potential impact of the next influenza pandemic
  on a national primary care medical workforce. Hum
  Resour Health 200537.
• 2. Sintchenko V, Gallego B. Laboratory-guided
  detection of disease outbreaksthree generations
  of surveillance systems. Arch Pathol Lab Med.
  2009 Jun133(6)916-25.

43
Discussion

• As for future research on information
  infrastructures for infectious disease response
  programs, studies are needed on
• - how evidence is defined and revised as new
  infectious diseases progress, and
• - how organizational and intellectual
  factors influence uptake of evidence in
  situations when the timeframe for taking
  preventive action is short (1).
• To achieve this, methods for evidence syntheses
  in the areas of outbreak detection and predictive
  modeling need to be established, including
  definition of criteria for evaluation of study
  quality.
• 1. Eccles MP, Armstrong D, Baker R, Cleary K,
  Davies H, Davies S, et al. An implementation
  research agenda. Implement Sci 2009418.

44
Conclusions

• The PROSPER protocol has been drafted for
  establishment of evidence-based pandemic response
  also in developing countries with limited access
  to advanced technology.
• The protocol is also useful because it
  facilitates the systematic study of the aspects
  of infrastructure and context that forms barriers
  to or facilitates response programs. In this way,
  existing and future information technologies can
  more effectively be summoned for analyses of new
  infectious diseases.
• It is necessary to establish consensus guidelines
  specifically for reporting of evidence derived
  from predictive modeling related to infectious
  disease.