| Notes |
epic.org EPIC-20-03-24-OSTP-FOIA-20200715-Production-pt1 000027
COVID-19 Tech Taskforce
Workstream Name: Clinical Health Data
Members: AWS, Color Genomics, ESRI, Gates, Google, Ro, NAVIGO
Goals and opportunities
• Assist researchers at the CDC, academic institutions, and other public-health officials to
capture clinical location data to assess resource capacity, effectiveness of policy
interventions, community education, and disease surveillance.
• Gather patient information (including behaviors, symptoms, and risk factors) in order to
create a model that can act as a high-value entry point.
• Assist health systems with de-identification and anonymization tasks.
• Understand and comply with relevant privacy and health regulation without impeding our
effectiveness.
• Clearly define the problem we are trying to solve so we can coordinate and execute as
quickly and effectively as possible.
• Leverage data to help prevent an unnecessary bottleneck on our medical resources.
• Clearly identify the symptoms of COVID-19 in particular geographies to promote
effective testing and healthcare response.
Planned operation outcome
Identify and provide clinical datasets for use in developing forecasting models and informing
disease surveillance.
Ideal operation outcomes [disclaimer: we don't know whether any/all of these are likely
or their feasibility; they're meant to be illustrative of the types of analyses possible with
broader clinical-data availability]
• Aggregate accurate information at federal, state, and county level.
• Develop forecast models predicting incidences and correlating them to key data points,
like number of available hospital beds and resource needs.
• Maximize limited testing resources by ensuring that the right groups of individuals are
tested. For instance, this could be determined based on the location data of individuals
who have tested positive.
• Along with location-based information, determine the cause of the spread of COVID-19
and mitigate continued spread. For instance, consider creating a rating system indicating
the danger of the spread of COVID-19 in certain communities.
• Track the different strains and expressions of the illness. There are multiple strains of
varying severity in Europe.
• Develop contact tracing solutions to help reduce spread at the top of the funnel. This
approach has been used successfully in other countries.
epic.org EPIC-20-03-24-OSTP-FOIA-20200715-Production-pt1 000028
• Follow interoperabil ity standards like FHIR to allow for easier collaboration .
• Understand comorbidity, including the symptoms displayed and how they manifest so
healthcare can improve outcomes .
• Determine whether task force resources are best spent assisting CDC and other
agencies with their own questions , or in posing our own.
• Reduce the likelihood of additional mutagenesis . This could include a review of the
potential presence of the virus in the United States at an earlier stage.
• Provide treatments tailored to specific strains of COVI D-19. Two key considerations:
o (1) Maintain a supply chain that accurately allocates emerging medicines;
o (2) Set clear guidelines and establish a decision tree reflecting the risk/reward of
particular kinds of treatments in different geographies .
• Understand interagency / intergovernmental collaboration and information sharing.
Other known working groups I efforts
• NAVIGO has a public-facing assessment that could be used for screening , education ,
data creation/collection , and other ideas to help improve the interaction at testing
centers. Prototype here:
• CT A's public link to telehealth companies questionna ire
• There is a substantial amount of interagency I intergovernmental work that is already
happening ; one of the challenges will be to ensure appropriate coordination among
government entities and the task force .
Concerns
• Working effectively while complying with regulation, including HIPAA and PHI, and
maintaining ethical data-sharing policies . Key considerations here will be understanding
who is using the data (e.g., is it open source) and what level of precision will be
provided.
• One meta-consideration is the government's tolerance for data aggregation /
centralizatio n. I.e., should we (1) gather large amou nts of data and allow for good
centralized governance to use that data to make decisions or (2) should we work to
narrow the scope of questions now so we don't expend time sorting through information
after we've gathered it?
2
epic.org EPIC-20-03-24-OSTP-FOIA-20200715-Production-pt1 000029
COVID-19 Tech Taskforce
Workstream Name: Hosting and Privacy
Members: AWS, Camber, ESRI, Facebook, Google, Harvard School of Public Health,
Microsoft, Mozilla, R4, SAP, Salesforce
Goals and opportunities
To assist researchers at the CDC, academic institutions, and other public-health officials,
capture non-clinical location data to assess contact rates, social distancing, effectiveness of
policy interventions, and disease surveillance.
Planned operation outcome
Create a secure, privacy respecting environment for researchers to aggregate and analyze data
to forecast and model COVID-19 related location and clinical data.
Existing data
At the moment, there are no publicly available, or aggregated datasets generally available to
researchers. Although some companies have aggregated data, the aggregated data are not
made publicly available and, at the moment, are limited only to a set of academic researchers
and epidemiologists. As of Tuesday, March 17, 2020, none of the members of the task force
have agreed to share data through the task force without further definition of requirements.
Concerns
The primary concerns related to privacy relate to understanding the extent of consent at the
point of data collection, considerations around deidentification and aggregation, guarding
against re-identification, maintaining clear guidelines around use of data (including the roles of
individuals accessing data, data storage/retention, and data reuse), and around transparency of
work. Additionally, security and compliance (both technical and regulatory) present challenges
for any data sharing effort.
Anticipated Deliverables
• A "Principles" document. We need a set of agreed-upon principles to help guide
technical and implementation details. This will guide all subsequent deliverables for the
task force.
• Assistance to entities around deidentification/anonymization efforts (provided sufficient
regulatory I legal approval/authority)
epic.org EPIC-20-03-24-OSTP-FOIA-20200715-Production-pt1 000030
COVID-19 Tech Taskforce
Workstream Name: Location Data
Members: Camber, ESRI, Facebook, Google, Harvard School of Public Health, Nielsen,
Unacast, Uber, Veraset
Goals and opportunities
To assist researchers at the CDC, academic institutions, and other public-health officials,
capture non-clinical location data to assess contact rates, social distancing, effectiveness of
policy interventions, and disease surveillance.
Planned operation outcome
Identify and provide non-clinical geolocation datasets for use in developing forecasting models
and informing disease surveillance and policy actions to prevent further spread of the disease.
We should leverage existing data to make mathematically informed decisions regarding where
to place resources.
Possible questions to be answered with our models
• Are social distancing policies being obeyed? Are they effective?
• Are people staying at home? How do we support them?
• What kind of specific messaging works and what doesn't to encourage desired
outcomes?
• What is causing spread in one area, and how does it allow us to mitigate the risk of
spread in other areas?
• What patterns are we seeing in transportation? Are people moving away from public
transportation in favor of private means?
• How can we use data collected to ensure we maintain a supply chain that allows the
necessary resources to be allocated to those communities that need them most (e.g.,
respirators, test kits, etc.)?
• How can we coordinate inbound requests to private industry for information to different
companies from different agencies and branches of government?
Possible products
What are the deliverables that government needs from us? We proposed the following:
• Development of a data pipeline
• Development of coordinated dashboards
o Ideally dashboards will be considered an authoritative source
o Information could be gathered from WHO, CDC, and foreign governments
epic.org EPIC-20-03-24-OSTP-FOIA-20200715-Production-pt1 000031
o Information gathered could include: number of individuals tested, confirmed
infected, still under investigation, school closures, unemployment, other business
metrics, etc.
Existing data
For this section, we have put together the high-level information we received on the call, please
add additional information or correct any errors . What data can we provide? What would be
useful for us to gather?
• Mobile location data (Camber/Veraset/Unacast)
• Trip data (Uber)
• Dashboards (ESRI)
o Data on hospital beds, number of beds (existing), staffed beds (supported), ICU
beds and bed utilization rates.
o https://coronavirus-resources .esri.com/
• Media behavior changes (Nielsen)
• Potential sociodemographic information and point of interest information
Other known working groups / efforts
Concerns
For this section, we have put together the high-level information we received on the call, please
add additional information or correct any errors .
• Engaging federal, state, and county government so we don't lose time in coordinating
• Complying with regulation and establishing the appropriate data aggregation threshold
o Clear messaging that data is aggregated and not personalized
o Clear about our data aggregation threshold. This will be different based on size
and density of community, and whether it's an open or closed dataset
• Privacy (see Hosting/Privacy workstream)
• Coordination
• Redundancy
• Sparsity of data/ data quality
• Pipeline access/aggregation/hosting
• Update frequency
• Ethics/punitive use of data
• Ensuring the models we build are effective in future, and not models of past. We need to
start predicting, or we won't be helpful in allowing gov to execute on plans and policies
for current use case of prevention
Requests to the government
2
epic.org EPIC-20-03-24-OSTP-FOIA-20200715-Production-pt1 000032
• Who do we coordinate with in government and relevant agencies (FEMA, Census, etc.)?
• What are the top questions that we can answer for government? This information will be
key in helping us inform our approach and get started more quickly
• Data use agreements
o Public access files, public utility files
• How should we coordinate with Federal, State, and local government most effectively?
3 |