Meeting Summaries of the CSA Expert Group on Modelling Approaches

As part of the response to the COVID-19 pandemic, at the request of Health Canada, Canada’s Chief Science Advisor (CSA) assembled the Expert Group on Modelling Approaches to obtain practical opinion from domain experts in order to inform her advice to the federal government regarding recovery strategies, data accessibility and data gaps.

On this page:

• Meeting #1 of the Expert Group on Modelling Approaches
• Meeting #2 of the Expert Group on Modelling Approaches
• Meeting #3 of the Expert Group on Modelling Approaches
• Meeting #4 of the Expert Group on Modelling Approaches
• Meeting #5 of the Expert Group on Modelling Approaches

The following discussion reflects evidence and scientific knowledge up to March 30.

Participants

• Mona Nemer, co-chair, Chief Science Advisor of Canada
• Stephen Lucas, co-chair, Deputy Minister Health Canada
• Nicole Basta, McGill University
• Caroline Colijn, Simon Fraser University
• Dan Coombs, University of British Columbia
• Jonathan Dushoff, McMaster University
• David Earn, McMaster University
• David Fisman, University of Toronto
• Seyed Moghadas, York University
• Nick Ogden, Public Health Agency of Canada
• Babak Pourbohloul, University of British Columbia
• Ashleigh Tuite, University of Toronto
• Jianhong Wu, York University

Guests

• Alejandro Adem, President, Natural Sciences and Engineering Research Council
• Ian Shugart, Clerk, PCO

Introduction

• This Expert Group is a sub-committee of the COVID19 Expert Panel.

Initial Questions for the Expert Group

From the Government perspective, modelling questions that could benefit from expert discussion include:

• What information do we need to enable better modelling?
• What can mathematical modellers provide?
  • Short term prediction of the epidemic curve (what assumptions are used and how different variables are inputted).
  • How well is social distancing working and how is it affecting the epidemic course?
  • How can we translate infection spread predictions into health system needs?
• How should we transition to new normal? (e.g. regional, sectors-schools/businesses considerations)
• What data, tools and infrastructure are needed to do the best possible job?

Experts initial discussions on the above and related questions:

Access to Data

• Ensuring that the modelling community has the data they need was identified as a top priority.
• Many existing models are based on data from international populations. Canadian data is required for responses tailored to Canada.
• More granular data is needed to strengthen modelling efforts, including: time of hospitalization, time in ICU, dates of discharge, detailed behavioural data, prevalence by age and location, serial data for longitudinal studies (not necessarily from same individual), combining telehealth data with hospital data.
• Ventilators: Dashboards indicating the percentage of ventilators that are in use (in a given hospital) for COVID-19 could identify underutilized resources.
• Expanding on the types of data collected could also help better understand the situation, e.g., testing people entering hospital for other reasons (heart attacks, motor accidents, births, etc.). This data could help estimate prevalence and inform infection control efforts.

Deployment of Technology

• DNA Testing: All existing data, including individuals who have been tested and those who have tested negative. Understanding why a person was tested, when and how many times will help. Australia is doing repeated testing on individuals who have been exposed. This can provide a better understanding of transmissibility.
• Serology: Large-scale serology is currently the largest gap. British Columbia is leading the way in collecting blood for serology, which could be important for determining who can go back to work.
• Mobile tracking: A number of apps are being developed/used across the world to better understand exposure, transmission and risk. Christophe Fraser, University of Oxford, wrote a speculative piece on how such an app could help control and slow the pandemic.
• Artificial Intelligence (AI): There is work underway to develop and deploy apps that reflect the Canadian context.

Efficacy of Mitigation Strategies

• Additional epidemiological studies are needed and are important for the collection of data. Adding to the granularity will be challenging in the current context given the already overburdened workload of those who collect the data.
• In the short term, estimating the impact of social distancing on the pandemic will depend on our ability to process good quality data in a timely way.
• It will be important to compare information across provinces to determine what works and what doesn’t. There are opportunities to learn from what other countries are doing.

Next Steps

• Health Canada will raise Identified data needs with the provinces.
• The Expert Group will meet again to expand on issues raised at the first meeting.

The following discussion reflects evidence and scientific knowledge up to April 9.

Participants

• Mona Nemer, co-chair, Chief Science Advisor of Canada
• Stephen Lucas, co-chair, Deputy Minister Health Canada
• Nicole Basta, McGill University
• Caroline Colijn, Simon Fraser University
• Dan Coombs, University of British Columbia
• Jonathan Dushoff, McMaster University
• David Earn, McMaster University
• David Fisman, University of Toronto
• Seyed Moghadas, York University
• Nick Ogden, Public Health Agency of Canada
• Ashleigh Tuite, University of Toronto
• Jianhong Wu, York University

Guests

• Alejandro Adem, President, Natural Sciences and Engineering Research Council

Localized Approaches

• Messaging should reflect that differences in trends across provinces are expected, and that general patterns are aligned. Aggregating data could be problematic by masking regional differences.
• Public health networks are talking across provinces about general guidelines.

Understanding Prevalence

• More needs to be done to understand prevalence, in particular broad and representative serological testing and surveillance.
  • Initially, will need to determine priorities for testing – where will most useful data come from. Data on children could be particularly useful.

Data Gaps

• Modellers are in need of more granular data.
• There are certain low cost solutions that could be addressed presently, e.g., Statistics Canada data. The Canadian mortality database could provide useful information, if it is collected in a timely manner.
• Granularity is not necessary for all individuals. A snapshot will do – a few hundred data points from key places.
• Modellers could benefit from more data on the impact of social distancing.

Modelling

• More complex modelling is needed to help think about how and when social distancing measures can be lessened. Comparing the different context across regions could provide insight, e.g. varying school shutdowns dates, when spring breaks occurred and implementation of provincial policies.
• Need to better understand the impact of lessening social distancing restrictions.

Next Steps

• The Expert Group on Modelling will develop priority recommendations. Short term and longer term recommendations can be provided in sequence.

The following discussion reflects evidence and scientific knowledge up to May 13.

Participants

• Mona Nemer, co-chair, Chief Science Advisor of Canada
• Caroline Colijn, Simon Fraser University
• Dan Coombs, University of British Columbia
• Jonathan Dushoff, McMaster University
• David Earn, McMaster University
• Seyed Moghadas, York University
• Nick Ogden, Public Health Agency of Canada
• Ashleigh Tuite, University of Toronto
• Jianhong Wu, York University

Guests

• Alejandro Adem, President, Natural Sciences and Engineering Research Council

Data and Modelling Challenges

• Modellers could benefit from aggregate data as well as data related to prevalence (serial estimates).
• Data on cases that are community transmissions versus outbreaks would be useful.
• De-escalation modelling requires pre-escalation data on age groups, professions, etc, through serology. However, de-escalations will have different conditions than before e.g., school re-openings in some regions. One week on, one week off, combined with guidelines, could be a good strategy for children returning to school.
• Actions that will be important going forward include contact tracing, ensuring people isolate when necessary, monitoring behaviour, and supporting people in isolation.
• There could be benefit in meeting with privacy commissioners with respect to data sharing.

Immunity Task Force

• The Expert Group on Modelling was asked for feedback on the Task Force’s Core Data Elements document. Comments were collected and shared with the Task Force secretariat.

Data Monitoring Initiative

• The expert group was asked for their thoughts on areas in need of new funding or infrastructure:
  • There is a need to generate more detailed data sets.
  • Solutions to easily and quickly access more data, while maintaining privacy.
  • We should ensure that we have individuals with the skills to manage databases to exploit data.
  • Develop metrics on how we are doing during de-escalation.
  • Accurately measuring R (measure of how fast virus is spreading) will be key during next phase.

Next Steps

• The Office of the Chief Science Advisor will coordinate a fourth meeting of the Expert Group on Modelling Approaches to discuss a strategy for moving forward and preparation for response to a second wave – what we should be monitoring and the information it requires (specific metrics).
  • The Expert Group will engage local health authorities with expertise in surveillance and contact tracing.

The following discussion reflects evidence and scientific knowledge up to May 31.

Participants

• Mona Nemer, co-chair, Chief Science Advisor of Canada
• Stephen Lucas, co-chair, Deputy Minister Health Canada
• Nicole Basta, McGill University
• Caroline Colijn, Simon Fraser University
• Dan Coombs, University of British Columbia
• Jonathan Dushoff, McMaster University
• David Earn, McMaster University
• David Fisman, University of Toronto
• Nick Ogden, Public Health Agency of Canada
• Ashleigh Tuite, University of Toronto

Guests

• Alejandro Adem, President, Natural Sciences and Engineering Research Council
• Sarah Collier, Manager, Surveillance and Epidemiology, Toronto Public Health
• Vera Etches, Chief Medical Officer, Ottawa

Surveillance Data

• There is an opportunity to make better use of existing data, such as negatives, asymptomatic cases and the locations of data points.
• There is a need for surveillance that can identify early outbreaks.
• Direct interaction with frontline public health is critical for modelling to identify challenges and gaps.
• As the economy reopens, the context in which data collection has occurred to this point will change. It will, therefore, be necessary to refocus attention on new contexts where transmission could be likely to occur.
• Local health authorities would like to extract additional data about workplaces such as types of infrastructures and industries associated with high risk of infection.
• COVID Near You (visualization to help identify hotspots) could be leveraged at the local level.
• Modellers could benefit from access to data on the number of cases per day and where they are occurring as well as how many in a distinct setting (e.g. community transmission versus superspreader location).
• In Ontario, there is data on the proportion of cases occurring outside of institutions, but additional data sources would help, e.g., a consistent survey across provinces. BC CDC has developed a standardized form.

Contact Tracing

• Processes differ from province to province – manual data entry results in longer processing times. Strengthening data management systems is key.
• Processes have improved across the information pipeline, however, manual processing during certain stages are still a challenge.
• Contact tracing teams have been expanding and are focused on training and data quality.
• There needs to be a quicker timeframe from symptom onset to contacting someone who tests positive in order to trace past interactions before recollection fades.
• Contact tracing efforts could benefit from stronger federal messaging: People need to 1) get tested when symptoms first occur, 2) self isolate until test results are in, 3) notify people they have come in contact with as soon as possible.
• More can be done to enable physicians to initiate contact tracing at diagnosis.

Next Steps

• Discuss forward strategy at next meeting, including further discussion on surveillance and contact tracing.
  • Develop list of things that modellers could do to help local health officials and data that could be useful for that.

The following discussion reflects evidence and scientific knowledge up to June 14.

Participants

• Mona Nemer, co-chair, Chief Science Advisor of Canada
• Stephen Lucas, co-chair, Deputy Minister Health Canada
• Nicole Basta, McGill University
• Caroline Colijn, Simon Fraser University
• Dan Coombs, University of British Columbia
• Jonathan Dushoff, McMaster University
• David Earn, McMaster University
• Ashleigh Tuite, University of Toronto
• Jianhong Wu, York University

Guests

• Sarah Collier, Manager, Surveillance and Epidemiology, Toronto Public Health

Local Health Authority Questions for Modellers

• Themes from questions from local health authority network included intervention, monitoring, effective workload for public health officials, details for “bubbles”( e.g., size, who, etc), schools, testing, superspreaders.
• It was highlighted by the group that promoting increased self-reporting could be a relatively quick and easy way to reduce spread.
• School settings expose many of the existing unknowns about the virus: transmission of small versus large droplets, transmission by paper or reduced transmission through handwashing, masks ,etc.
  • Developing web interface software could provide opportunity for modellers to collaborate by focussing on different parameters, however, this would require a large effort and would need to be done in careful and close interaction between modellers and decision makers to ensure results are interpreted properly.
  • Development of a flow chart is another options, e.g., illustrate the different elements of a model (data vs assumptions).
  • The flow chart could be complemented with a list of key questions that are useful at a particular stage.
  • This effort could also link to modelling that’s already been done, creating transparency and interoperability.
• Local health officials could benefit from being connected to the broader modelling network.

Follow-up:

• OCSA will cluster the topics from public health officials’ network in categories and will recirculate to the group and PHAC to develop a list of experts for each category.

Impact of COVID-19 Response Measures

• Sorting out one impact measure from another, e.g., masks vs social distancing, would be challenging. Because so many measures have been put in place at the same time, it would be hard to extract specific measures – difficult to understand what the curve would look like if mask wearing was or was not introduced in a given context.
• Going forward, it would help to survey the public about hand washing, mask use, etc. The City of Ottawa has started doing this.
• Researchers in Canada are looking at the effectiveness of N95 versus surgical masks, however, there are many challenges to studying the impact of masks (homemade/surgical/N95) vs no-masks in context such as schools or public places.

Identifying Early Outbreaks

• Continuing surveillance and contact tracing will be critical.
• Would be important for the data reporting to convey to public the category and individual is in.

Follow-up:

• OCSA will check-in with PHAC on efforts to leverage wastewater infrastructure for COVID-19 response and follow up with expert group.

Next Steps

• Follow-up call to discuss ways in which modellers could make predictions about transmission in schools. Expert Group members will provide recommendations on experts on this topic who should be invited to the call.